WO2018019210A1 - 一种监测交互式网络电视iptv用户状态的方法及装置 - Google Patents
一种监测交互式网络电视iptv用户状态的方法及装置 Download PDFInfo
- Publication number
- WO2018019210A1 WO2018019210A1 PCT/CN2017/094151 CN2017094151W WO2018019210A1 WO 2018019210 A1 WO2018019210 A1 WO 2018019210A1 CN 2017094151 W CN2017094151 W CN 2017094151W WO 2018019210 A1 WO2018019210 A1 WO 2018019210A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- quality difference
- user
- indicator
- difference
- quality
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 77
- 238000012544 monitoring process Methods 0.000 title claims abstract description 16
- 238000012216 screening Methods 0.000 claims abstract description 22
- 239000011159 matrix material Substances 0.000 claims description 163
- 238000000556 factor analysis Methods 0.000 claims description 26
- 230000002452 interceptive effect Effects 0.000 claims description 14
- 230000001186 cumulative effect Effects 0.000 claims description 7
- 230000004048 modification Effects 0.000 claims description 7
- 238000012986 modification Methods 0.000 claims description 7
- 238000012217 deletion Methods 0.000 claims description 4
- 230000037430 deletion Effects 0.000 claims description 4
- 230000005484 gravity Effects 0.000 claims 1
- 238000005457 optimization Methods 0.000 abstract description 7
- 230000008447 perception Effects 0.000 abstract description 2
- 239000000523 sample Substances 0.000 description 16
- 230000006866 deterioration Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 7
- 238000001914 filtration Methods 0.000 description 4
- 238000007405 data analysis Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000013011 mating Effects 0.000 description 2
- 238000009877 rendering Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000000275 quality assurance Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44213—Monitoring of end-user related data
- H04N21/44218—Detecting physical presence or behaviour of the user, e.g. using sensors to detect if the user is leaving the room or changes his face expression during a TV program
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N17/00—Diagnosis, testing or measuring for television systems or their details
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N17/00—Diagnosis, testing or measuring for television systems or their details
- H04N17/002—Diagnosis, testing or measuring for television systems or their details for television cameras
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44204—Monitoring of content usage, e.g. the number of times a movie has been viewed, copied or the amount which has been watched
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44213—Monitoring of end-user related data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/60—Network structure or processes for video distribution between server and client or between remote clients; Control signalling between clients, server and network components; Transmission of management data between server and client, e.g. sending from server to client commands for recording incoming content stream; Communication details between server and client
- H04N21/63—Control signaling related to video distribution between client, server and network components; Network processes for video distribution between server and clients or between remote clients, e.g. transmitting basic layer and enhancement layers over different transmission paths, setting up a peer-to-peer communication via Internet between remote STB's; Communication protocols; Addressing
- H04N21/643—Communication protocols
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/60—Network structure or processes for video distribution between server and client or between remote clients; Control signalling between clients, server and network components; Transmission of management data between server and client, e.g. sending from server to client commands for recording incoming content stream; Communication details between server and client
- H04N21/63—Control signaling related to video distribution between client, server and network components; Network processes for video distribution between server and clients or between remote clients, e.g. transmitting basic layer and enhancement layers over different transmission paths, setting up a peer-to-peer communication via Internet between remote STB's; Communication protocols; Addressing
- H04N21/643—Communication protocols
- H04N21/64322—IP
Definitions
- the present invention relates to the technical field of interactive network television data analysis and processing, and in particular, to a method and apparatus for monitoring an interactive network television IPTV user status.
- IPTV Internet Protocol Television
- the method of processing and analyzing IPTV data is largely judged according to the value model of MOS (Mean Opinion Score) designed by major set-top box manufacturers, but there is no uniform standard between the manufacturers models, and the result is difficult to verify. Therefore, for the operator, it is currently impossible to determine the status of each user in the IPTV system in a timely and accurate manner, so that when a user with perceived deterioration appears in the IPTV system, it is difficult to optimize the network of the IPTV system in time and affect the user experience.
- MOS Mean Opinion Score
- An object of the embodiments of the present invention is to provide a method and apparatus for monitoring an IPTV user state of an interactive network television, which enables an operator to determine the status of each user in the IPTV system in a timely and accurate manner, so that the perceived deterioration occurs in the IPTV system.
- the network optimization of the IPTV system is carried out in time to enhance the user experience.
- an embodiment of the present invention provides a method for monitoring an IPTV user state of an interactive network television, including:
- program viewing data of each IPTV user includes a plurality of viewing records, each of the viewing records including a plurality of first numerical indicators;
- the status of each IPTV user is determined based on the quality difference records in the plurality of viewing records of each IPTV user.
- the step of filtering out the quality difference records in the plurality of viewing records of each IPTV user according to the plurality of first numerical indicators in each of the viewing records of each IPTV user includes:
- the viewing record of the IPTV user matches the threshold of the quality difference indicator in the quality difference recording model, the viewing record is determined to be a quality difference record.
- Q' 1 denotes the first quality difference index in the quality difference recording model
- ⁇ 1 denotes the threshold value of the first quality difference index
- Q' 2 denotes the second quality difference index in the quality difference recording model
- ⁇ 2 denotes The threshold of the second qualitative difference indicator
- Q′ i represents the i-th qualitative difference indicator in the quality difference recording model
- ⁇ i represents the threshold of the i-th qualitative difference indicator
- Q′ represents the quality difference indicator in the qualitative difference recording model.
- Quantity Quantity.
- the step of determining the status of each IPTV user according to the quality difference record in the plurality of viewing records of each IPTV user includes:
- f 2 represents a quality difference user model
- D represents the number of viewing records of the IPTV user
- d i represents a mark of the quality difference record distribution of the ith user's ith viewing record, a threshold indicating the proportion of the difference record to the number of viewing records
- the step of determining, according to the quality difference record in the plurality of viewing records of the IPTV user, the marking of the quality difference record distribution of each viewing record of the IPTV user including:
- the method further includes: before the step of filtering the quality difference records in the plurality of viewing records of each IPTV user according to the plurality of first numerical indicators in each of the viewing records of each IPTV user, the method further includes:
- Obtaining a threshold value of the quality difference indicator and the quality difference indicator in the quality difference recording model according to a plurality of second numerical type indicators included in the viewing record of the plurality of qualitative difference users obtained in advance, and a quality difference record in the quality difference user model The threshold of the proportion of the number of viewing records.
- the threshold values of the quality difference indicator and the quality difference indicator in the quality difference recording model are acquired according to the plurality of second numerical indicators included in the viewing records of the plurality of qualitative users obtained in advance, and the quality in the quality difference user model
- the step of recording the difference between the difference record and the number of the recorded records includes:
- the threshold values of the quality difference index and the quality difference index in the quality difference record model are obtained, and the threshold value of the quality difference record in the quality difference user model accounts for the proportion of the number of the viewing records.
- the step of obtaining an indicator correlation matrix of each pre-determined user of the quality difference according to the plurality of second numerical indicators of each of the pre-acquired quality users includes:
- Standardizing a plurality of second numerical indicators of each of the previously obtained quality users Obtaining a plurality of standardized numerical indicators of each of the previously obtained quality users;
- a correlation between each of the two normalized numerical indicators of each of the previously obtained qualitative users is calculated, and an index correlation matrix of each of the previously obtained qualitative users is obtained.
- the step of normalizing a plurality of second numerical indicators of each of the previously obtained quality users to obtain a plurality of standardized numerical indicators of each of the previously obtained quality users includes:
- the step of calculating the correlation between each of the two standardized numerical indicators of each of the previously obtained qualitative users, and obtaining the index correlation matrix of each of the previously obtained qualitative users includes:
- the indicator correlation matrix and the preset clustering The quantity, the step of determining the second numerical indicator included in the indicator variable of each cluster of each of the previously obtained quality users, including:
- the R-type clustering method is used to determine the index variable included in each cluster of each pre-determined user of the difference according to the similarity distance matrix of each pre-determined user and the preset number of clusters.
- the second numerical indicator is used to determine the index variable included in each cluster of each pre-determined user of the difference according to the similarity distance matrix of each pre-determined user and the preset number of clusters.
- the similarity distance between each two second numerical indicators of each of the previously obtained qualitative users is calculated according to the index correlation matrix of each of the previously obtained qualitative users, and each previously obtained quality difference is obtained.
- the steps of the user's similarity distance matrix including:
- the similarity distance matrix of the previously obtained quality difference user is calculated; where S represents the similarity distance matrix of the previously obtained quality difference user, and Q n ' represents the number of the second numerical type fingers.
- the method also includes:
- the clustering result pedigree of each pre-existing user is drawn, and the clustering result pedigree is displayed.
- determining, in each of the pre-determined quality users, the steps in the indicator correlation matrix are different from each other, including:
- the multiple rows are deleted to one row according to the deletion instruction received on the operation interface, and the second numerical indicator corresponding to the deleted row is deleted, so that the indicator correlation is performed.
- the rows in the matrix are different from each other.
- the method further includes:
- the distance matrix of each pre-obtained user of the difference is calculated according to the index correlation matrix of each pre-determined user of the difference, and each distance matrix of each user is obtained according to the distance matrix of each user.
- Multi-dimensional scale map of pre-acquired quality users is calculated according to the index correlation matrix of each pre-determined user of the difference, and each distance matrix of each user is obtained according to the distance matrix of each user.
- the multidimensional scale map of the drawing is presented, and the second numerical indicator included in the indicator variable of each cluster of each pre-determined user of the difference is modified according to the modification instruction received at the operation interface.
- the multi-dimensional scaling method is used to calculate a distance matrix of each pre-determined user of the quality difference according to the index correlation matrix of each of the previously obtained qualitative users, including:
- the step of filtering out representative indicators of each cluster of each of the previously obtained qualitative users from the second numerical indicators included in the indicator variables of each cluster of each of the previously obtained qualitative users include:
- the factor analysis method is used to obtain an elementary load matrix of the index variable of each cluster of each pre-determined user of the difference according to a factor analysis model of each cluster of index variables of each of the pre-determined users;
- the variance contribution of each common factor in the model is analyzed according to the factor of the index variable of each cluster of each of the previously obtained qualitative users, and the rotation of the indicator variable of each cluster of each of the previously obtained qualitative users a load factor in the elementary load matrix, and a correlation contribution model is established for each cluster of indicator variables of each of the pre-determined users of the difference;
- a correlation contribution model is established according to each of the clustered index variables of each of the previously obtained qualitative users, and a second numerical indicator included in each cluster of index variables included in each of the previously obtained qualitative users is selected.
- a second numerical indicator having the highest relevance contribution is used, and the second numerical indicator is used as a representative index of the cluster.
- the variance contribution degree of each common factor in the model is analyzed according to the factor of the index variable of each cluster of each pre-determined user, and the index variable of each cluster of each pre-determined user of the difference.
- the load factor in the rotated elementary load matrix, the step of establishing a correlation contribution model for each of the clustered indicator variables of each of the pre-determined users of the difference including:
- the correlation variable model is established by calculating the index variable of each cluster of the user who obtains the quality difference in advance; wherein RC t represents the index variable of the cluster to establish a correlation contribution model.
- the step of obtaining a threshold value of the quality difference indicator and the quality difference indicator in the quality difference record model according to the selected representative index, and a threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records includes:
- Each second is counted according to the representative index of each cluster of each pre-existing quality user Numerical indicators are screened as the number of times the indicator is represented;
- the second numerical indicators that are selected as representative indicators are sorted according to the order of the number of times;
- the threshold value of the quality difference index in the quality difference recording model and the threshold value of the quality difference record in the quality difference user model accounted for the number of the viewing records are obtained.
- the step of obtaining a threshold value of the quality difference indicator in the quality difference record model according to the selected quality difference indicator, and a threshold value of the ratio of the quality difference record in the quality difference user model to the number of the viewing records includes:
- the threshold value of the quality difference index in the quality difference recording model is set as a first preset value, and each of the quality difference recording indicators and the threshold value of the quality difference index set as the first preset value is used to determine each a mark of the distribution of the quality difference records of each of the viewing records of the pre-obtained quality users;
- Controlling the threshold value of the quality difference indicator as the first preset value of the quality difference recording model, and the threshold value of the difference of the number of the difference records in the number of the viewing records is set as the second preset value of the quality difference user model from the plurality of first users and / or a plurality of second users are selected as the quality difference user; wherein the type of the first user is a non-quality user, and the type of the second user is a quality user;
- a preset value of the quality difference recording model, and a threshold value of the proportion of the difference record to the number of the viewing records is set as the adjusted second preset value of the quality difference user model screening quality user accuracy reaches a third preset value And adjusting the adjusted first preset value as a threshold value of the quality difference indicator, and using the adjusted second preset value as a threshold value of the proportion of the quality difference record to the number of the viewing records.
- the quality difference record of the viewing record is recorded.
- the distribution flag is set to 1;
- the quality difference of the viewing record is The flag of the record distribution is set to zero.
- the step of determining that each of the pre-obtained users of the quality difference is greater than or equal to the second preset value includes:
- the first preset value is adjusted until the ratio of the previously obtained quality difference user is greater than or equal to the second preset value.
- An embodiment of the present invention further provides an apparatus for monitoring an IPTV user status of an interactive network television, including:
- a first acquiring module configured to acquire program viewing data of each IPTV user; wherein the program viewing data includes a plurality of viewing records, each of the viewing records includes a plurality of first numerical indicators;
- a screening module configured to filter out the quality difference records in the plurality of viewing records of each IPTV user according to the plurality of first numerical indicators in each viewing record of each IPTV user;
- Determining a module set to record a quality difference in a plurality of viewing records for each IPTV user, Determine the status of each IPTV user.
- a computer storage medium is further provided, and the computer storage medium may store an execution instruction for performing the implementation of the method for monitoring an interactive network television IPTV user state in the foregoing embodiment.
- the quality difference records in the plurality of viewing records of each IPTV user are filtered according to the first numerical indicator in the plurality of viewing records of each IPTV user, and the quality difference is selected according to the screening. Recording, determining the status of each IPTV user, and solving the problem that the operator cannot determine the status of each user in the IPTV system in a timely and accurate manner, so that it is difficult to timely perform network on the IPTV system when a user with perceived deterioration appears in the IPTV system.
- FIG. 1 is a flowchart of a method for monitoring an IPTV user state of an interactive network television according to a first embodiment of the present invention
- FIG. 2 is a flow chart showing threshold values of a quality difference index and a quality difference index in a quality difference recording model and a threshold value of a quality difference record in a user model of a quality difference in the number of viewing records in the first embodiment
- FIG. 3 is a flowchart of performing refined analysis on a user with a quality difference in a specific example in the first embodiment of the present invention
- FIG. 5 is a multi-dimensional scale diagram drawn in a specific example in the first embodiment of the present invention.
- FIG. 6 is a schematic structural diagram of an apparatus for monitoring an IPTV user state of an interactive network television according to a second embodiment of the present invention.
- FIG. 7 is a schematic diagram of an IPTV data analysis architecture in a third embodiment of the present invention.
- a first embodiment of the present invention provides a method for monitoring an IPTV user state of an interactive network television, the method comprising:
- Step 101 Acquire program viewing data of each IPTV user.
- the program viewing data includes a plurality of viewing records, each of the viewing records includes a plurality of first numerical indicators, and each of the first numerical indicators is different from each other.
- the first numerical indicator may be avg-bit-rate, multi-abend-numbers, delay (df), and unicast buffer underflow (vod-).
- Step 102 Filter out the quality difference records in the plurality of viewing records of each IPTV user according to the plurality of first numerical indicators in each of the viewing records of each IPTV user.
- the above-mentioned quality difference recording refers to a viewing record when the user feels that the program quality is deteriorated (for example, a card, etc.) during the process of watching the program.
- screening may be performed by detecting whether the values of the plurality of first numerical indicators in each of the viewing records of the IPTV user match the thresholds of the quality difference indicators in the quality difference recording model.
- a quality record in a plurality of viewing records for each IPTV user Specifically, if it is detected that the value of the plurality of first numerical indicators in the viewing record of the IPTV user matches the threshold of the quality difference indicator in the quality difference recording model, the viewing record is determined to be a quality difference record.
- Q' 1 denotes the first quality difference index in the quality difference recording model
- ⁇ 1 denotes the threshold value of the first quality difference index
- Q' 2 denotes the second quality difference index in the quality difference recording model
- ⁇ 2 denotes The threshold of the second qualitative difference indicator
- Q′ i represents the i-th qualitative difference indicator in the quality difference recording model
- ⁇ i represents the threshold of the i-th qualitative difference indicator
- Q′ represents the quality difference indicator in the qualitative difference recording model.
- the quantity, F refers to the relationship in which the quality difference indicators are arranged in combination.
- the quality difference indicators in the quality difference recording model are different from each other, and similar to the first numerical type index, may be avg-bit-rate, multicast buffer underflow times (multi-abend) -numbers), delay (df), unicast buffer underflow (vod-abend-numbers), packet loss factor (mlr), number of requests (req-numbers), jitter (jitter), switching time (acc-avr -time), can-use-rate, total number of unicast application failures (vod-fail-numbers), number of underflows (abend-numbers), total number of playback errors (play-error-numbers), Play-time or overflow-numbers.
- the quality difference recording model includes two quality difference indicators, namely jitter (represented by Q' 1 ) and delay (represented by Q' 2 ), respectively.
- the recording model Q' 1 > ⁇ 1 indicates that the jitter is greater than 4
- Q' 2 > ⁇ 2 indicates that the delay is greater than 10
- the jitter of the plurality of first numerical indicators in the IPTV user's viewing record is 5 If the value of the delay is 12, the watch record is considered to be a quality record.
- Step 103 Determine the status of each IPTV user according to the quality difference record in the plurality of viewing records of each IPTV user.
- the values of the first numerical indicators are matched with the thresholds of the qualitative difference indicators in the qualitative difference recording model, and D i ⁇ f 1 represents the values of the plurality of first numerical indicators of the i-th viewing record and the qualitative difference recording model
- the threshold of the quality difference indicator does not match; then the mark of the distribution is recorded according to the quality difference of each viewing record of the IPTV user, and the user model is passed through the quality difference.
- f 2 represents a quality difference user model
- D represents the number of viewing records of the IPTV user
- a threshold value indicating the proportion of the quality difference record to the number of viewing records, wherein if the value of f 2 is 1, it is determined that the IPTV user is a quality user. If the value of f 2 is 0, it is determined that the IPTV user is a non-quality user.
- the quality difference user refers to the IPTV user who feels the deterioration of the program quality (such as the Karton screen) during the process of watching the program.
- the non-quality user refers to the feeling that the program quality is excellent during the process of watching the program. IPTV users.
- the determined quality difference user may be a quality user of the Karton screen, and may of course be other types of quality users.
- the type of user whose quality is determined depends mainly on the type of each quality difference indicator in the quality difference record model. Specifically, if each of the quality difference indicators in the quality difference recording model is used to judge the quality difference user of the Katton flower screen, then the determined quality difference user type is the quality difference user of the Karton flower screen, similarly, if the quality difference is Each quality difference indicator in the record model is used to judge other types of quality difference users, and then the determined quality difference user type is other types of quality difference users.
- the quality difference records in the plurality of viewing records of each IPTV user are filtered according to the first numerical indicator in the plurality of viewing records of each IPTV user, and are filtered according to the selected
- the quality record records the status of each IPTV user, which solves the problem that the operator cannot determine the status of each user in the IPTV system in a timely and accurate manner, so that it is difficult for the IPTV system to be timely in the case of users with perceived deterioration in the IPTV system.
- the problem of network optimization and affecting the user experience is to enable the operator to determine the status of each user in the IPTV system in a timely and accurate manner, and to optimize the network of the IPTV system in time when the user with perceived deterioration appears in the IPTV system. To enhance the user experience.
- the method before the step 102, further includes: acquiring the quality difference record model according to the plurality of second numerical indicators included in the viewing records of the plurality of qualitative users obtained in advance The threshold of the quality difference indicator and the quality difference indicator in the medium, and the threshold of the proportion of the quality difference record in the quality difference user model to the number of the viewing records.
- the threshold values of the quality difference indicator and the quality difference indicator in the quality difference recording model are acquired according to a plurality of second numerical type indicators included in the viewing records of the plurality of qualitative difference users obtained in advance, and the quality difference user is obtained.
- a step package of the threshold value of the difference in the number of viewing records in the model include:
- Step 201 Obtain an indicator correlation matrix of each of the previously obtained quality difference users according to a plurality of second numerical indicators of each of the previously obtained quality difference users.
- the plurality of second numerical indicators are different from each other, and similar to the first numerical indicator, may be avg-bit-rate, multi-abend-numbers, and delay.
- Time (df) unicast buffer underflow (vod-abend-numbers), packet loss factor (mlr), number of requests (req-numbers), jitter (jitter), switching time (acc-avr-time), available Rate (can-use-rate), total number of unicast application failures (vod-fail-numbers), number of underflows (abend-numbers), total number of playback errors (play-error-numbers), playback duration (play- Time) or overflow-numbers, etc.
- Step 202 Determine, according to the index correlation matrix of each pre-determined quality user and the preset number of clusters, a second numerical value included in an indicator variable of each cluster of each pre-determined user of the difference index.
- Step 203 Filter out representative indicators of each cluster of each of the previously obtained qualitative users from the second numerical indicators included in the indicator variables of each cluster of each of the previously obtained qualitative users.
- Step 204 Obtain a threshold value of the quality difference indicator and the quality difference indicator in the quality difference record model according to the selected representative index, and a threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records.
- the step 201 includes the following steps:
- a plurality of second numerical indicators of each of the previously obtained quality users are standardized, and a plurality of standardized numerical indicators of each of the previously obtained quality users are obtained.
- z dq represents a standardized numerical indicator of the qth second numerical indicator of the dth viewing record
- z dq represents the dth viewing
- the qth second numerical indicator recorded Indicates the sample mean of the qth second numerical indicator
- s q represents the sample standard deviation of the qth second numerical indicator
- D n represents the number of the nth pre-obtained qualitative user's viewing record
- Q represents the The dimension of the two numerical indicators.
- the correlation between each of the two standardized numerical indicators of each of the previously obtained qualitative users is calculated, and an index correlation matrix of each of the previously obtained qualitative users is obtained.
- R represents the index correlation matrix of the user of the quality difference obtained in advance
- r ij represents the correlation between the i-th normalized numerical index and the j-th normalized numerical index
- z di represents the first record of the d-th record a standardized numerical indicator of i second numerical indicators
- z dj represents a standardized numerical indicator of the j-th second numerical indicator of the d-th viewing record, Indicates the sample mean of the jth second numerical indicator.
- the step 202 includes the following steps:
- each row in the indicator correlation matrix of each of the previously obtained qualitative users is different from each other.
- the specific implementation manner of the first step is: determining whether there are multiple rows in the indicator correlation matrix (ie, determining whether there are duplicate rows), and if there are multiple rows in the indicator correlation matrix, according to
- the deletion instruction received on the operation interface deletes multiple rows to one row, and deletes the second numerical indicator corresponding to the deleted row, so that the rows in the indicator correlation matrix are different from each other.
- the delete command can be input by the administrator on the operation interface according to his own experience, which directly indicates which rows are deleted.
- the similarity distance between each two second numerical indicators of each pre-determined user is calculated according to the index correlation matrix of each of the previously obtained qualitative users, and each pre-obtained The similarity distance matrix of the quality user.
- the user's similarity distance matrix, Q n ' represents the number of second numerical type fingers.
- the R-type clustering method is used to determine the index of each cluster of each pre-determined user of the difference according to the similarity distance matrix of each pre-determined user and the preset number of clusters.
- the second numerical indicator contained in the variable is used to determine the index of each cluster of each pre-determined user of the difference according to the similarity distance matrix of each pre-determined user and the preset number of clusters.
- each of the previously obtained quality difference users may be determined according to the similarity distance matrix of each pre-determined user of the difference and the preset number of clusters by the R-type squared sum (Ward) clustering method.
- the second numerical indicator included in the indicator variable for each cluster may be determined according to the similarity distance matrix of each pre-determined user of the difference and the preset number of clusters by the R-type squared sum (Ward) clustering method.
- the second numerical indicator included in the indicator variable for each cluster may be determined according to the similarity distance matrix of each pre-determined user of the difference and the preset number of clusters by the R-type squared sum (Ward) clustering method.
- the second numerical indicator included in the indicator variable for each cluster may be determined according to the similarity distance matrix of each pre-determined user of the difference and the preset number of clusters by the R-type squared sum (Ward) clustering method.
- the Ward method includes the formula for calculating the squared and distance of the deviation in the class and the formula for calculating the squared distance between the classes.
- the formula for calculating the squared distance of the deviation in the class is:
- the indicator variable of the cluster can be expressed as X t represents the index variable of the t-th cluster, T represents the number of clusters, and z 1 represents the first second-value indicator of the indicator variables of the t-th cluster, L t of the second numerical indicators Indicator variable represents the t-th cluster in.
- the method further includes: each of the users according to each of the previously obtained quality differences The second numerical indicator included in the index variable of the cluster, the clustering result pedigree map of each pre-existing qualitative user is drawn, and the steps of the clustering result pedigree map are displayed, so that the administrator can quickly, It is clear to know the second numerical indicator included in the indicator variable of each cluster of each of the previously obtained qualitative users.
- the method further includes the following steps:
- the multi-dimensional scaling method is used to calculate the distance matrix of each pre-determined user of the difference according to the index correlation matrix of each pre-existing user of the difference, and the distance of the user according to each pre-determined quality difference A matrix that plots the multidimensional scale of each pre-fetched user.
- the multidimensional scale map is drawn, and the second numerical indicator included in the indicator variable of each cluster of each pre-determined user is modified according to the modification instruction received in the operation interface.
- the administrator determines, according to the multi-dimensional scale map, the second item included in the indicator variable of each cluster of each (or some) previously obtained quality difference user determined If the numerical indicator is inaccurate, the administrator can modify the second numerical indicator included in the indicator variable of each cluster of each (or some) pre-determined user of the difference by inputting the modification instruction in the operation interface. .
- step 203 specifically includes the following steps:
- a factor analysis model is established for each cluster of indicator variables of each of the previously obtained qualitative users.
- the factor analysis model of the indicator variable of each cluster among them F 1 , F 2 ,..., F m are common factors, and their coefficients ⁇ qm are called load factors, which represent the correlation coefficient between the qth second numerical index and the mth factor, and ⁇ q is a special factor.
- each factor of the user is obtained according to each pre-determined quality difference by a factor analysis method.
- the factor analysis model of the index variable of the class obtains the elementary load matrix of the index variable of each cluster of each pre-determined user of the difference.
- the specific implementation manner of the second step is: first calculating a correlation coefficient matrix R t between each of the second numerical indicators in each cluster, and then calculating the characteristic value of R t Corresponding feature vector
- calculating ⁇ 1 represents the elementary load matrix
- each pre-determined user of the difference is determined according to the threshold of the cumulative variance contribution rate of the preset feature root and the elementary load matrix of the index variable of each cluster of each pre-determined user of the difference The number of common factors for each cluster of indicator variables.
- the smallest integer is the number of common factors u(u ⁇ L t ), where ⁇ represents the threshold of the cumulative variance contribution rate of the preset eigenvalue.
- the fourth step orthogonally rotating the elementary load matrix of the index variable of each cluster of each of the previously obtained qualitative users, and rotating the index variable of each cluster according to each of the previously obtained qualitative users
- the variance contribution of each common factor in the factor analysis model of the index variable of each cluster of each pre-determined user is calculated.
- ⁇ 1 (u) is ⁇ 1 before u column
- T is positive
- the mating matrix, the load factor after rotation is Then pass the formula
- Con bt represents the variance contribution of the bth common factor.
- the variance contribution of each common factor in the model is analyzed according to the factor of the index variable of each cluster of each pre-determined user, and the index of each cluster of each pre-determined user of the difference
- the load factor in the elementary load matrix after the rotation of the variable establishes a correlation contribution model for each cluster of indicator variables of each of the previously obtained qualitative users.
- the correlation variable model is established by calculating the index variable of each cluster of the previously obtained quality difference user, wherein RC tq represents the correlation contribution model of the qth second numerical indicator of the t-th cluster, L t represents the number of second numerical indicators in the index variable of the cluster, u represents the number of common factors of the index variable of the cluster, and b represents the serial number of the common factor,
- the load factor in the rotated elementary load matrix of the index variable representing the cluster Con bt represents the variance contribution of the bth common factor of the index variable of the cluster
- T represents the number of clusters, and RC t represents the index of the cluster
- the variables establish a correlation contribution model.
- the correlation contribution model is established according to the index variable of each cluster of each pre-determined user of the difference, and the second value included in the indicator variable of each cluster of each pre-determined user of the difference is obtained.
- the second numerical indicator with the highest relevance contribution is selected in the type index, and the second numerical indicator is used as the representative index of the cluster.
- the step 204 includes the following steps:
- the number of times each of the second numerical indicators is screened as the representative index is counted according to the representative index of each cluster of each of the previously obtained qualitative users.
- the second numerical indicators that are selected as representative indicators are sorted according to the order of the number of times.
- the quality difference indicator in the quality difference record model is selected from the second numerical indicator that is filtered to represent the indicator.
- the administrator can input a selection instruction according to his own experience, and the selection instruction instructs selection of a quality difference indicator that can reflect the network condition.
- the threshold value of the quality difference index in the quality difference record model is obtained, and the threshold value of the quality difference record in the quality difference user model accounts for the proportion of the number of the viewing records.
- the specific implementation of the fourth step is:
- the threshold of the quality difference indicator in the quality difference recording model is set as a first preset value (wherein, if the quality difference recording model includes a plurality of quality difference indicators, each of the quality difference indicators corresponds to a first preset value And each preset value may be different from each other, and each pre-determined quality difference user is determined according to the chromatographic record model formed by the selected quality difference indicator and the threshold value of the quality difference indicator set as the first preset value.
- the flag of the quality difference record distribution of the viewing record is set to 1; if the value of the second numerical type indicator in the viewing record of the user of the quality difference obtained in advance, and the quality difference record model If the threshold of the quality difference indicator set as the first preset value does not match, the flag of the quality difference record distribution of the viewing record is set to 0;
- the number of the quality difference records of each of the previously obtained quality difference users is counted, and according to each of the previously obtained quality difference users The number of the quality difference records, and the ratio of the quality difference records of each of the previously obtained quality difference users to the number of the viewing records of each of the previously obtained quality difference users is calculated;
- the threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records is set to a second preset value, and the ratio of each of the previously obtained quality difference users is determined to be greater than or equal to the second preset value.
- the step of determining that the ratio of each of the previously obtained quality difference users is greater than or equal to the second preset value comprises: determining whether the ratio of the previously obtained quality difference user is greater than or equal to the second preset value; If the ratio of the user is less than the second preset value, the first preset value is adjusted until the ratio of the previously obtained quality difference user is greater than or equal to the second preset value.
- the threshold value of the threshold value of the control quality index is the first preset value
- the threshold value of the difference of the number of the quality records to the number of the viewing records is set as the second preset value of the quality difference user model from the plurality of first
- the user and/or the plurality of second users are selected as the quality difference user; wherein the type of the first user is a non-quality user, and the type of the second user is a quality user;
- the threshold value of the threshold value of the first difference is obtained, and the threshold value of the proportion of the difference record in the number of the viewing records is set as the second preset value, and the user model is selected to screen the user of the quality difference.
- Accuracy rate if the accuracy reaches the third preset value (for example, 70%), the first preset value is used as the threshold value of the quality difference index, and the second preset value is used as the ratio of the quality difference record to the number of the viewing records.
- the size of the second preset value is adjusted, and the first preset value is adjusted according to the adjusted second preset value until the threshold of the quality difference indicator is adjusted.
- the first preset value of the quality difference recording model, and the threshold value of the proportion of the difference record to the number of the viewing records is set as the adjusted second preset value of the quality difference user model screening quality user accuracy rate reaches And a third preset value, and the adjusted first preset value is used as a threshold value of the quality difference indicator, and the adjusted second preset value is used as a threshold value of the proportion of the quality difference record in the number of the viewing records.
- the threshold value of the quality difference index and the quality difference index in the obtained quality difference record model is further explained by a specific example, and the proportion of the quality difference record in the quality difference user model accounts for the number of the viewing records.
- the step of the threshold is further explained by a specific example, and the proportion of the quality difference record in the quality difference user model accounts for the number of the viewing records.
- the user of the quality of the Karton screen is taken as an example, and a user of the quality of the Karton screen is used as an example for detailed analysis, wherein the user of the quality difference has 32 viewing records, and the user of the quality difference
- the second numerical indicator has a dimension of 23.
- the steps to perform a refined analysis of the user of the quality difference are as follows:
- Step 301 Obtain an indicator correlation matrix according to a plurality of second numerical indicators of the quality difference user.
- the plurality of second numerical indicators are different from each other, and similar to the first numerical indicator, may be avg-bit-rate, multi-abend-numbers, and delay.
- Time (df) unicast buffer underflow (vod-abend-numbers), packet loss factor (mlr), number of requests (req-numbers), jitter (jitter), switching time (acc-avr-time), available Rate (can-use-rate), total number of unicast application failures (vod-fail-numbers), number of underflows (abend-numbers), total number of playback errors (play-error-numbers), playback duration (play- Time) or overflow-numbers, etc.
- Step 302 Determine, according to the indicator correlation matrix and the preset number of clusters, a second numerical indicator included in the indicator variable of each cluster.
- Step 303 Filter out representative indicators of each cluster from the second numerical indicators included in the indicator variables of each cluster.
- Step 304 Obtain a threshold value of the quality difference indicator and the quality difference indicator in the quality difference record model according to the selected representative index, and a threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records.
- step 301 specifically includes the following steps:
- the first step is to standardize a plurality of second numerical indicators to obtain a plurality of standardized numbers. Value indicator.
- z dq represents a standardized numerical indicator of the qth second numerical indicator of the dth viewing record
- z dq represents the qth of the dth viewing record
- the second numerical indicator Indicates the sample mean of the qth second numerical indicator
- s q represents the sample standard deviation of the qth second numerical indicator
- D n represents the number of viewing records of the qualitative user
- the correlation between each of the two standardized numerical indicators is calculated, and the index correlation matrix is obtained.
- R is the index correlation of the user of the quality difference
- r ij represents the correlation between the i-th normalized numerical index and the j-th normalized numerical index
- z di represents the standardized numerical index of the i-th second numerical indicator of the d-th viewing record
- z di represents a standardized numerical indicator of the j-th second numerical indicator of the d-th viewing record, Indicates the sample mean of the jth second numerical indicator.
- the step 302 includes the following steps:
- the first step it is determined that the rows in the indicator correlation matrix of the user of the difference are different from each other.
- the specific implementation manner of the first step is: determining whether there are multiple rows in the indicator correlation matrix (ie, determining whether there are duplicate rows), and if there are multiple rows in the indicator correlation matrix, according to
- the delete instruction received on the operation interface deletes multiple rows to one row, and deletes the second numerical indicator corresponding to the deleted row, so that the rows in the indicator correlation matrix are mutually Not the same.
- the delete command can be input by the administrator on the operation interface according to his own experience, which directly indicates which rows are deleted.
- the similarity distance between each two second numerical indicators of the user of the difference is calculated, and the similarity distance matrix of the user of the difference is obtained.
- the second numerical indicator included in the indicator variable of each cluster of the qualitative user is determined by the R-type clustering method according to the similarity distance matrix of the user of the difference and the preset number of clusters.
- each of the clusters of the qualitative user may be determined according to the similarity distance matrix of the user of the difference and the preset number of clusters (for example, 5) by the R-type squared sum (Ward) clustering method.
- the second numerical indicator included in the indicator variable may be determined according to the similarity distance matrix of the user of the difference and the preset number of clusters (for example, 5) by the R-type squared sum (Ward) clustering method.
- the second numerical indicator included in the indicator variable included in the indicator variable.
- the Ward method includes the formula for calculating the squared and distance of the deviation in the class and the formula for calculating the squared distance between the classes.
- the formula for calculating the squared distance of the deviation in the class is:
- the method further includes: drawing a clustering result pedigree diagram of the qualitative user according to the second numerical indicator included in the indicator variable of each cluster of the qualitative user, and displaying the clustering result of the drawing
- the steps of the pedigree diagram so that the administrator can quickly and clearly know the second numerical indicator included in the indicator variable of each cluster of the qualitative user.
- the clustering result pedigree diagram is shown in Fig. 4.
- 1 in the abscissa axis represents avg-bit-rate
- 2 represents df
- 3 represents multi-abend-numbers
- 4 represents jitter
- 5 represents can.
- abend-numbers, vod-abend-numbers, multi-abend-numbers can be divided into the first category; play-error-numbers, mlr can be divided into the second category; avg-bit-rate , df, jitter can be divided into the third category; overflow-numbers can be divided into the fourth category, req-numbers and other remaining second numerical indicators can be divided into the fifth category.
- the method further includes the following steps:
- the multidimensional scaling method is used to calculate the distance matrix of the user of the quality difference according to the index correlation matrix of the user of the difference, and the multidimensional scale map of the user of the quality difference is drawn according to the distance matrix of the user of the difference.
- the multidimensional scale map is drawn, and the second numerical indicator included in the indicator variable of each cluster of the qualitative user is modified according to the modification instruction received in the operation interface.
- the administrator determines the previous multi-dimensional scale map, If the second numerical indicator included in the indicator variable of each cluster of the user of the quality difference is inaccurate, the administrator can modify the index variable of each cluster of the qualitative user by inputting the modification instruction in the operation interface.
- the second numerical indicator included in the figure If the administrator determines the previous multi-dimensional scale map, If the second numerical indicator included in the indicator variable of each cluster of the user of the quality difference is inaccurate, the administrator can modify the index variable of each cluster of the qualitative user by inputting the modification instruction in the operation interface.
- the second numerical indicator included in the figure is the administrator determines the previous multi-dimensional scale map, If the second numerical indicator included in the indicator variable of each cluster of the user of the quality difference is inaccurate, the administrator can modify the index variable of each cluster of the qualitative user by inputting the modification instruction in the operation interface. The second numerical indicator included in the figure.
- step 303 specifically includes the following steps:
- a factor analysis model is established for the indicator variables of each cluster of the qualitative user.
- the factor analysis model of the indicator variable of each cluster among them F 1 , F 2 ,..., F m are common factors, and their coefficients ⁇ qm are called load factors, which represent the correlation coefficient between the qth second numerical index and the mth factor, and ⁇ q is a special factor.
- the elementary load matrix of the index variable of each cluster of the qualitative user is obtained.
- the specific implementation manner of the second step is: first calculating a correlation coefficient matrix R t between each of the second numerical indicators in each cluster, and then calculating the characteristic value of R t Corresponding feature vector
- calculating ⁇ 1 represents the elementary load matrix
- the publicity of the index variable of each cluster of the qualitative user is determined. The number of factors.
- the elementary load matrix of the index variable of each cluster of the qualitative user is orthogonally rotated, and is calculated according to the load factor in the rotated elementary load matrix of the index variable of each cluster of the user of the quality difference.
- the variance contribution of each common factor in the factor analysis model of the indicator variable of each cluster of the qualitative user is calculated according to the load factor in the rotated elementary load matrix of the index variable of each cluster of the user of the quality difference.
- ⁇ 1 (u) is ⁇ 1 before u column
- T is positive
- the mating matrix, the load factor after rotation is Then pass the formula
- Con bt represents the variance contribution of the bth common factor.
- the variance contribution of each common factor in the model is analyzed according to the factor of the index variable of each cluster of the user of the quality difference, and the rotation of the elementary load matrix of each cluster of the qualitative difference user
- the load factor establishes a correlation contribution model for the indicator variables of each cluster of the qualitative user.
- the indicator variable of each cluster of the user with poor quality is established to establish a correlation contribution model, followed by the formula
- the correlation variable model is established by calculating the index variable of each cluster of the qualitative user, wherein RC tq represents the correlation contribution model of the qth second numerical indicator of the t-th cluster, and L t represents poly The number of the second numerical indicator in the index variable of the class, u represents the number of common factors of the index variable of the cluster, and b represents the serial number of the common factor.
- Con bt represents the variance contribution of the bth common factor of the index variable of the cluster
- T represents the number of clusters
- RC t represents the index of the cluster
- the correlation contribution model is established according to the index variable of each cluster of the user of the qualitative difference, and the correlation contribution is selected from the second numerical indicator included in the indicator variable of each cluster of the qualitative user.
- the second numerical indicator is used as the representative index of the cluster.
- the step 304 includes the following steps:
- the number of times each second numerical indicator is screened as the representative index is counted.
- the second numerical indicators that are selected as representative indicators are sorted according to the order of the number of times.
- the quality difference indicator in the quality difference record model is selected from the second numerical indicator that is filtered to represent the indicator.
- the administrator can input a selection instruction according to his own experience, and the selection instruction indicates the selection.
- a quality indicator that reflects the state of the network.
- the threshold value of the quality difference index in the quality difference record model is obtained, and the threshold value of the quality difference record in the quality difference user model accounts for the proportion of the number of the viewing records.
- the specific implementation of the fourth step is:
- the threshold of the quality difference indicator in the quality difference recording model is set as a first preset value (wherein, if the quality difference recording model includes a plurality of quality difference indicators, each of the quality difference indicators corresponds to a first preset value And each preset value may be different from each other, and each of the viewing records of the user of the quality difference is determined according to the qualitative difference record model of the selected quality difference indicator and the threshold value of the quality difference indicator set as the first preset value.
- the mark of the distribution of the quality difference records is set as a first preset value (wherein, if the quality difference recording model includes a plurality of quality difference indicators, each of the quality difference indicators corresponds to a first preset value And each preset value may be different from each other, and each of the viewing records of the user of the quality difference is determined according to the qualitative difference record model of the selected quality difference indicator and the threshold value of the quality difference indicator set as the first preset value.
- the quality difference record distribution of the viewing record is The flag is set to 1; if the value of the second numerical indicator in the viewing record of the quality difference user does not match the threshold of the quality difference indicator set as the first preset value in the quality difference recording model, the viewing record is The mark of the quality difference record distribution is set to 0;
- the number of the quality difference records of the user of the quality difference is counted, and the quality difference record of the user of the quality difference is calculated according to the number of the quality difference records of the user of the quality difference.
- the threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records is set to a second preset value, and the ratio of the quality difference user is determined to be greater than or equal to the second preset value.
- the step of determining that the ratio of the user of the quality difference is greater than or equal to the second preset value comprises: determining whether the ratio of the user of the quality difference is greater than or equal to the second preset value; if the ratio of the user of the quality difference is less than the second preset value, Adjust the first preset value until the ratio of the user of the quality difference is greater than or equal to the second preset value.
- the threshold value of the threshold value of the control quality index is the first preset value
- the threshold value of the difference of the number of the quality records to the number of the viewing records is set as the second preset value of the quality difference user model from the plurality of first
- the user and/or the plurality of second users are selected as the quality difference user; wherein the type of the first user is a non-quality user, and the type of the second user is a quality user;
- the threshold value of the threshold value of the first difference is obtained, and the threshold value of the ratio of the difference record to the number of the viewing records is set as the second preset value. Selecting the accuracy rate of the user with poor quality. If the accuracy reaches the third preset value (for example, 70%), the first preset value is used as the threshold of the quality difference indicator, and the second preset value is recorded as the quality difference record.
- the threshold of the proportion of the recorded quantity is obtained
- the accuracy rate does not reach the third preset value
- the size of the second preset value is adjusted, and the first preset value is adjusted according to the adjusted second preset value until the threshold of the quality difference indicator is adjusted.
- the first preset value of the quality difference record model, and the threshold value of the difference between the quality record and the number of the watch record is set as the adjusted second preset value of the quality difference user model screening quality user accuracy reaches the third pre- A value is set, and the adjusted first preset value is used as a threshold value of the quality difference index, and the adjusted second preset value is used as a threshold value of the proportion of the quality difference record to the number of the viewing records.
- the false positive rate is only 6.8%, but the quality difference of 59 aggravate screens is The user's prediction accuracy is as high as 66.1%.
- the method of the present invention not only achieves joint clustering of high-dimensional indicators of user viewing record data through multidimensional scaling method and R-type Ward clustering method, but also selects representative indicators from indicators through reverse factor analysis to achieve effective Dimensionality reduction.
- the quality difference record model and the quality difference user model are established according to the combination of the quality difference indicators, and the parameters of the quality difference record model and the quality difference user model are continuously iteratively optimized (ie, the quality difference record model)
- the threshold of the quality difference index and the quality difference index, and the threshold value of the difference value of the quality difference record in the user model of the quality difference account for the proportion of the number of viewing records), so that the accuracy rate and the false positive rate satisfy the requirement.
- a second embodiment of the present invention provides an apparatus for monitoring an IPTV user status of an interactive network television, including:
- the first obtaining module 601 is configured to acquire program viewing data of each IPTV user; wherein the program viewing data includes a plurality of viewing records, each of the viewing records includes a plurality of first numerical indicators;
- the screening module 602 is configured to: screen, according to the plurality of first numerical indicators in each of the viewing records of each IPTV user, the quality difference records in the plurality of viewing records of each IPTV user;
- the determining module 603 is set to be based on the quality difference in the plurality of viewing records of each IPTV user Record and determine the status of each IPTV user.
- the screening module 602 includes:
- the first screening unit is configured to detect whether the value of the plurality of first numerical indicators in each of the viewing records of the IPTV user matches the threshold of the quality difference indicator in the quality difference recording model, and if the IPTV user views the record The value of the plurality of first numerical indicators matches the threshold of the qualitative difference indicator in the qualitative difference recording model, and triggers the second screening unit;
- the second screening unit is configured to determine that the viewing record is a quality difference record according to a trigger of the first screening unit.
- Q' 1 denotes the first quality difference index in the quality difference recording model
- ⁇ 1 denotes the threshold value of the first quality difference index
- Q' 2 denotes the second quality difference index in the quality difference recording model
- ⁇ 2 denotes The threshold of the second qualitative difference indicator
- Q′ i represents the i-th qualitative difference indicator in the quality difference recording model
- ⁇ i represents the threshold of the i-th qualitative difference indicator
- Q′ represents the quality difference indicator in the qualitative difference recording model.
- Quantity Quantity.
- the determining module 603 includes:
- a third screening unit configured to determine, according to the quality difference record in the plurality of viewing records of the IPTV user, a flag of a quality difference record distribution of each viewing record of the IPTV user;
- a fourth screening unit configured to mark the distribution of the quality difference records according to each viewing record of the IPTV user, and pass the quality difference user model Determining the value of f 2 of the IPTV user; wherein f 2 represents a quality difference user model, D represents the number of viewing records of the IPTV user, and d i represents a mark of the quality difference record distribution of the ith user's ith viewing record, a threshold indicating the proportion of the difference record to the number of viewing records;
- the fifth filtering means if the value is set to f 2. 1, it is determined that the user is a poor quality of IPTV user; if f is 02, it is determined that the user is a non poor quality IPTV user.
- the device further comprises:
- the second obtaining module is configured to obtain, according to the plurality of second numerical indicators included in the viewing records of the plurality of qualitative users obtained in advance, the thresholds of the quality difference indicators and the quality difference indicators in the quality difference recording model, and the quality difference The threshold of the proportion of the difference record in the user model to the number of viewing records.
- the function of the second obtaining module may be implemented by an IPTV analysis system.
- the second obtaining module includes:
- a first acquiring unit configured to obtain, according to each of the plurality of second numerical indicators of each of the pre-determined quality users, an indicator correlation matrix of each of the previously obtained qualitative users;
- a second obtaining unit configured to determine, according to the index correlation matrix of each of the previously obtained qualitative users and the preset number of clusters, the index variables included in each cluster of each of the previously obtained qualitative users The second numerical indicator;
- a third obtaining unit configured to filter out each cluster of each pre-determined user of the difference from the second numerical indicator included in the index variable of each cluster of each of the pre-determined users of the quality difference Representative indicator
- the fourth obtaining unit is configured to obtain a threshold value of the quality difference indicator and the quality difference indicator in the quality difference record model according to the selected representative index, and a threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records .
- the first obtaining unit includes:
- a first obtaining subunit configured to perform normalization processing on each of the plurality of second numerical indicators of each of the previously obtained quality difference users, to obtain a plurality of standardized numerical type indicators of each of the previously obtained quality difference users;
- the second obtaining subunit is configured to calculate a correlation between each of the two normalized numerical indicators of each of the previously obtained qualitative users, and obtain an index correlation matrix of each of the previously obtained qualitative users.
- the second acquisition subunit is also set to pass the formula Calculating the correlation between each of the two normalized numerical indicators of each of the previously obtained qualitative users; wherein r ij represents the correlation between the i-th normalized numerical indicator and the j-th normalized numerical indicator, z di represents a standardized numerical indicator of the i-th second numerical indicator of the d-th viewing record, a sample mean value representing the i-th second numerical indicator, and z dj represents a standardized numerical indicator of the j-th second numerical indicator of the d-th viewing record, a sample mean value representing the jth second numerical indicator;
- the second obtaining unit includes:
- a third obtaining subunit configured to determine that each row in the indicator correlation matrix of each of the previously obtained qualitative users is different from each other;
- the fourth obtaining subunit is set to be based on the index correlation of each pre-acquired quality user a matrix, calculating a similarity distance between each two second numerical indicators of each of the previously obtained qualitative users, and obtaining a similarity distance matrix of each of the previously obtained qualitative users;
- the fifth obtaining subunit is configured to determine, by the R-type clustering method, each of the pre-obtained users of the difference according to the similarity distance matrix of each pre-determined user of the difference and the preset number of clusters The second numerical indicator included in the clustered indicator variable.
- the fourth acquisition subunit is also set to pass the formula
- the similarity distance matrix of the previously obtained quality difference user is calculated; where S represents the similarity distance matrix of the previously obtained quality difference user, and Q n ' represents the number of the second numerical type fingers.
- the device further comprises:
- a first drawing module configured to map a clustering result pedigree of each of the previously obtained qualitative users according to a second numerical indicator included in an indicator variable of each cluster of each of the previously obtained qualitative users;
- a clustering result pedigree diagram showing the rendering.
- the third obtaining subunit is further configured to determine whether the same plurality of rows exist in the index correlation matrix, and if the same plurality of rows exist in the index correlation matrix, according to the deletion instruction received on the operation interface, The rows are deleted to one row, and the second numerical indicator corresponding to the deleted row is deleted, so that the rows in the indicator correlation matrix are different from each other.
- the device further comprises:
- a second drawing module is configured to calculate, by using a multi-dimensional scaling method, a distance matrix of each pre-determined user of the difference according to the index correlation matrix of each of the previously obtained qualitative users, and according to each pre-determined quality difference The distance matrix of the user, and draw a multidimensional scale map of each user who has obtained the quality difference in advance;
- Modifying the module setting to display the multi-dimensional scale map of the drawing, and modifying the index variable of each cluster of each pre-determined user of the difference according to the modification instruction received in the operation interface The second numerical indicator.
- the second drawing module includes:
- the first drawing unit is set to pass the formula Calculating the distance between each of the two second numerical indicators; wherein h ij represents the distance between the i-th second numerical indicator and the j-th second numerical indicator;
- the third obtaining unit includes:
- a sixth obtaining subunit configured to establish a factor analysis model for each of the clustered index variables of each of the previously obtained qualitative users
- a seventh obtaining subunit configured to obtain, by a factor analysis method, a factor analysis model of each cluster of indicator variables of each cluster of pre-determined users, to obtain each cluster of each pre-determined user of the difference The elementary load matrix of the indicator variable;
- the eighth obtaining subunit is configured to determine, according to a threshold value of the cumulative variance contribution rate of the preset feature root, and an elementary load matrix of the index variable of each cluster of each of the previously obtained qualitative users, The number of common factors of the indicator variable for each cluster of the user;
- a ninth acquisition subunit configured to orthogonally rotate an elementary load matrix of an indicator variable of each cluster of each of the previously obtained quality users, and according to each of the pre-determined quality differences of each user of the cluster Calculating the variance contribution of each common factor in the factor analysis model of the index variable of each cluster of each of the previously obtained qualitative users by the load factor in the rotated elementary load matrix of the index variable;
- a tenth obtaining subunit configured to analyze a variance contribution degree of each common factor in the model according to a factor of each index variable of each cluster of the user of each pre-determined quality difference, and each of the pre-obtained quality difference users a load factor in the rotated elementary load matrix of the clustered indicator variable, and a correlation contribution model is established for each clustered indicator variable of each of the previously obtained qualitative users;
- the eleventh acquisition subunit is set to each cluster of users according to each pre-determined quality difference
- the indicator variable establishes a relevance contribution model, and selects a second numerical indicator with the highest relevance contribution from the second numerical indicator included in each cluster of index variables of each pre-existing qualitative user, and The second numerical indicator is used as a representative index of the cluster.
- the tenth acquisition subunit is also set to pass the formula Obtaining a correlation contribution model for each cluster of index variables of each of the previously obtained qualitative users; wherein RC tq represents a correlation contribution model of the qth second numerical indicator of the t-th cluster, L t represents the number of second numerical indicators in the index variable of the cluster, u represents the number of common factors of the index variable of the cluster, and b represents the serial number of the common factor.
- the load factor in the rotated elementary load matrix of the index variable representing the cluster, Con bt represents the variance contribution of the bth common factor of the index variable of the cluster, and T represents the number of clusters;
- the tenth acquisition subunit is also set to pass the formula
- the correlation variable model is established by calculating the index variable of each cluster of the user who obtains the quality difference in advance; wherein RC t represents the index variable of the cluster to establish a correlation contribution model.
- the fourth obtaining unit includes:
- the twelfth obtaining subunit is configured to count, according to the representative index of each cluster of each pre-determined quality difference user, the number of times each second numerical indicator is selected as the representative index;
- the thirteenth obtaining subunit is configured to sort the second numerical indicators that are selected as representative indicators according to the number of times obtained by statistics, in descending order of the number of times;
- a fourteenth obtaining subunit configured to screen the quality difference indicator in the quality difference recording model from the second numerical indicator that is filtered to represent the indicator according to the selection instruction received on the operation interface;
- the fifteenth obtaining subunit is configured to obtain a threshold value of the quality difference index in the quality difference recording model according to the selected quality difference index, and a threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records.
- the fifteenth obtaining subunit is further configured to set a threshold value of the quality difference index in the quality difference recording model to a first preset value, and according to the selected quality difference indicator and the quality difference indicator set as the first preset value
- a quality difference recording model composed of thresholds, each record of each pre-determined user of the quality difference is determined a mark of the distribution of the quality difference records;
- the fifteenth obtaining subunit is further configured to count, according to the mark of the quality difference record distribution of each of the viewing records of each of the pre-acquisition users, the number of the quality difference records of each of the previously obtained quality users, and according to each a pre-obtained number of quality difference records of the user of the quality difference, and calculating a ratio of the quality difference records of each of the previously obtained quality difference users to the number of the viewing records of each of the previously obtained quality difference users;
- the fifteenth obtaining subunit is further configured to set a threshold value of the proportion of the quality difference record in the quality difference user model to the number of the viewing records as a second preset value, and determine that the ratio of each of the previously obtained quality difference users is greater than Or equal to the second preset value;
- the fifteenth obtaining subunit is further configured to control the quality difference recording model in which the threshold value of the quality difference index is the first preset value, and the threshold value of the proportion of the quality difference record in the number of the viewing records is set as the second preset value
- the difference user model selects the quality difference user from the plurality of first users and/or the plurality of second users; wherein the type of the first user is a non-quality user, and the type of the second user is a quality user;
- the fifteenth obtaining subunit is further configured to obtain a quality difference recording model in which the threshold value of the quality difference index is the first preset value, and a threshold value of the proportion of the quality difference record in the number of the viewing records is set as the second preset value
- the difference user model filters the accuracy of the user with the difference in quality. If the accuracy reaches the third preset value, the first preset value is used as the threshold of the quality difference indicator, and the second preset value is recorded as the quality difference record.
- the threshold of the proportion of the quantity is the first preset value, and a threshold value of the proportion of the quality difference record in the number of the viewing records.
- the fifteenth obtaining sub-unit is further configured to adjust the size of the second preset value if the accuracy rate does not reach the third preset value, and adjust the first preset value according to the adjusted second preset value, until
- the threshold value of the quality difference index is a quality difference recording model of the adjusted first preset value, and the threshold value of the proportion of the quality difference record to the number of the viewing records is set as the adjusted second user value of the quality difference user model screening quality
- the accuracy of the difference user reaches a third preset value, and the adjusted first preset value is used as the threshold of the quality difference indicator, and the adjusted second preset value is used as the proportion of the quality record of the difference record. Threshold.
- the fifteenth obtaining subunit is further configured to: if the value of the second numerical indicator in the viewing record of the user of the quality difference obtained in advance, and the quality difference indicator set as the first preset value in the quality difference recording model If the threshold is matched, the flag of the quality difference record distribution of the viewing record is set to 1;
- the fifteenth obtaining subunit is further set to a value of the second numerical indicator in the viewing record of the quality difference user obtained in advance, and a threshold value of the quality difference indicator set as the first preset value in the qualitative difference recording model is not If it matches, the flag of the quality difference record distribution of the viewing record is set to zero.
- the fifteenth obtaining subunit is further configured to determine whether the ratio of the previously obtained quality difference user is greater than or equal to the second preset value, and if the ratio of the previously obtained quality difference user is less than the second preset value, adjust the first The preset value is until the ratio of the previously obtained quality difference user is greater than or equal to the second preset value.
- the quality difference records in the plurality of viewing records of each IPTV user are selected according to the first numerical indicator in the plurality of viewing records of each IPTV user, and are filtered according to the selected
- the quality record records the status of each IPTV user, which solves the problem that the operator cannot determine the status of each user in the IPTV system in a timely and accurate manner, so that it is difficult for the IPTV system to be timely in the case of users with perceived deterioration in the IPTV system.
- the problem of network optimization and affecting the user experience is to enable the operator to determine the status of each user in the IPTV system in a timely and accurate manner, and to optimize the network of the IPTV system in time when the user with perceived deterioration appears in the IPTV system. To enhance the user experience.
- the apparatus for monitoring the state of the interactive network television IPTV user provided by the second embodiment of the present invention is the apparatus for applying the foregoing method for monitoring the state of the interactive network television IPTV user, that is, all the embodiments of the foregoing method are applicable to the method. Devices, and all achieve the same or similar benefits.
- the third embodiment of the present invention provides an IPTV data analysis architecture, including: a data acquisition module 701, a probe module 702, an IPTV service quality assurance system (IQAS), and an IPTV analysis system 704.
- the data acquisition module 701 is configured to capture a network packet when the user is watching the program, which can be implemented by a libpacp module.
- the probe module 702 is configured to parse the captured network packet and report it to the IQAS 703 for IPTV analysis.
- System 704 can The data of the user of the quality difference is obtained from the IQAS 703 for analysis, and the threshold values of the quality difference index and the quality difference index in the quality difference record model and the threshold value of the difference record of the quality difference record in the user model of the quality difference account for the number of the viewing records are determined.
- the quality difference records in the plurality of viewing records of each IPTV user are filtered according to the first numerical indicator in the plurality of viewing records of each IPTV user, and are filtered according to the selected
- the quality record records the status of each IPTV user, which solves the problem that the operator cannot determine the status of each user in the IPTV system in a timely and accurate manner, so that it is difficult for the IPTV system to be timely in the case of users with perceived deterioration in the IPTV system.
- the problem of network optimization and affecting the user experience is to enable the operator to determine the status of each user in the IPTV system in a timely and accurate manner, and to optimize the network of the IPTV system in time when the user with perceived deterioration appears in the IPTV system. .
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Social Psychology (AREA)
- Databases & Information Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Biomedical Technology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
本发明的实施例提供了一种监测交互式网络电视IPTV用户状态的方法及装置,其中,该方法包括:获取各IPTV用户的节目观看数据;其中,节目观看数据包括多条观看记录,每条观看记录包括多个第一数值型指标;根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录;根据每个IPTV用户的多条观看记录中的质差记录,确定出每个IPTV用户的状态。本发明的实施例能使运营商及时、准确的确定出IPTV***中的各用户的状态,使得在IPTV***中出现感知恶化的用户时,及时对IPTV***进行网络优化,提升用户体验。
Description
本发明涉及交互式网络电视数据分析与处理的技术领域,特别涉及一种监测交互式网络电视IPTV用户状态的方法及装置。
随着国内固网运营商的业务转型,交互式网络电视(IPTV,Internet Protocol Television)业务已呈现出快速增长的态势。尽管网络运营商也在重点采集、监测网络传输参数,但是IPTV业务在网络传输中仍会受到干扰,产生的不利影响主要表现在用户感知恶化,例如卡顿花屏等。
目前,处理分析IPTV数据的方法很大程度上是依据各大机顶盒产商设计的(MOS,Mean Opinion Score)值模型进行评判,但各产商模型之间缺乏统一标准,结果难以验证。因此,对于运营商而言,目前无法及时、准确的确定出IPTV***中的各用户的状态,使得在IPTV***中出现感知恶化的用户时,难以及时对IPTV***进行网络优化,影响用户体验。
发明内容
本发明实施例的目的在于提供一种监测交互式网络电视IPTV用户状态的方法及装置,能使运营商及时、准确的确定出IPTV***中的各用户的状态,使得在IPTV***中出现感知恶化的用户时,及时对IPTV***进行网络优化,提升用户体验。
为了达到上述目的,本发明的实施例提供了一种监测交互式网络电视IPTV用户状态的方法,包括:
获取各IPTV用户的节目观看数据;其中,节目观看数据包括多条观看记录,每条观看记录包括多个第一数值型指标;
根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选
出每个IPTV用户的多条观看记录中的质差记录;
根据每个IPTV用户的多条观看记录中的质差记录,确定出每个IPTV用户的状态。
其中,根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录的步骤,包括:
检测IPTV用户的每条观看记录中的多个第一数值型指标的数值,是否与质差记录模型中的质差指标的阈值匹配;
若IPTV用户的观看记录中的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值匹配,则确定该观看记录为质差记录。
其中,质差记录模型为f1=F(Q′1>φ1,Q′2>φ2,...,Q′i>φi),i=Q′,其中,f1表示质差记录模型,Q′1表示质差记录模型中的第一个质差指标,φ1表示第一个质差指标的阈值,Q′2表示质差记录模型中的第二个质差指标,φ2表示第二个质差指标的阈值,Q′i表示质差记录模型中的第i个质差指标,φi表示第i个质差指标的阈值,Q′表示质差记录模型中质差指标的数量。
其中,根据每个IPTV用户的多条观看记录中的质差记录,确定出每个IPTV用户的状态的步骤,包括:
根据IPTV用户的多条观看记录中的质差记录,确定IPTV用户的每条观看记录的质差记录分布的标记;
根据IPTV用户的每条观看记录的质差记录分布的标记,通过质差用户模型确定出IPTV用户的f2的值;其中,f2表示质差用户模型,D表示IPTV用户的观看记录的数量,di表示IPTV用户的第i条观看记录的质差记录分布的标记,表示质差记录占观看记录的数量的比重的阈值;
若f2的值为1,则确定该IPTV用户为质差用户;
若f2的值为0,则确定该IPTV用户为非质差用户。
其中,根据IPTV用户的多条观看记录中的质差记录,确定IPTV用户的每条观看记录的质差记录分布的标记的步骤,包括:
通过公式确定IPTV用户的每条观看记录的质差记录分布的标记;其中,Di=f1表示第i条观看记录的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值匹配,Di≠f1表示第i条观看记录的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值不匹配。
其中,在根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录的步骤之前,方法还包括:
根据预先得到的多个质差用户的观看记录所包含的多个第二数值型指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,根据预先得到的多个质差用户的观看记录所包含的多个第二数值型指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤,包括:
根据每个预先得到的质差用户的多个第二数值型指标,得到每个预先得到的质差用户的指标相关性矩阵;
根据每个预先得到的质差用户的指标相关性矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标;
从每个预先得到的质差用户的每个聚类的指标变量所包含的第二数值型指标中,筛选出每个预先得到的质差用户的每个聚类的代表指标;
根据筛选出的代表指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,根据每个预先得到的质差用户的多个第二数值型指标,得到每个预先得到的质差用户的指标相关性矩阵的步骤,包括:
对每个预先得到的质差用户的多个第二数值型指标进行标准化处理,
得到每个预先得到的质差用户的多个标准化数值型指标;
计算每个预先得到的质差用户的每两个标准化数值型指标之间的相关性,得到每个预先得到的质差用户的指标相关性矩阵。
其中,对每个预先得到的质差用户的多个第二数值型指标进行标准化处理,得到每个预先得到的质差用户的多个标准化数值型指标的步骤,包括:
通过公式计算得到每个预先得到的质差用户的多个标准化数值型指标;其中,zdq表示第d条观看记录的第q个第二数值型指标的标准化数值型指标,zdq表示第d条观看记录的第q个第二数值型指标,表示第q个第二数值型指标的样本均值,sq表示第q个第二数值型指标的样本标准差,Dn表示第n个预先得到的质差用户的观看记录的数量,Q表示第二数值型指标的维度。
其中,计算每个预先得到的质差用户的每两个标准化数值型指标之间的相关性,得到每个预先得到的质差用户的指标相关性矩阵的步骤,包括:
通过公式计算得到每个预先得到的质差用户的每两个标准化数值型指标之间的相关性;其中,rij表示第i个标准化数值型指标与第j个标准化数值型指标之间的相关性,zdi表示第d条观看记录的第i个第二数值型指标的标准化数值型指标,表示第i个第二数值型指标的样本均值,zdj表示第d条观看记录的第j个第二数值型指标的标准化数值型指标,表示第j个第二数值型指标的样本均值;
通过公式R=(rij)计算得到每个预先得到的质差用户的指标相关性矩阵;其中,R表示预先得到的质差用户的指标相关性矩阵。
其中,根据每个预先得到的质差用户的指标相关性矩阵和预设的聚类
数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标的步骤,包括:
确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同;
根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的每两个第二数值型指标之间的相似性距离,得到每个预先得到的质差用户的相似性距离矩阵;
通过R型聚类法,根据每个预先得到的质差用户的相似性距离矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
其中,根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的每两个第二数值型指标之间的相似性距离,得到每个预先得到的质差用户的相似性距离矩阵的步骤,包括:
通过公式sij=1-rij计算得到每两个第二数值型指标之间的相似性距离;其中,sij表示第i个第二数值型指标与第j个第二数值型指标之间的相似性距离;
其中,方法还包括:
根据每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标,绘制每个预先得到的质差用户的聚类结果谱系图,并展现绘制的聚类结果谱系图。
其中,确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同的步骤,包括:
判断指标相关性矩阵中是否存在相同的多个行;
若指标相关性矩阵中存在相同的多个行,则根据在操作界面接收到的删除指令,将多个行删除至一个行,并删除被删除行对应的第二数值型指标,使指标相关性矩阵中各行互不相同。
其中,在确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同的步骤之后,方法还包括:
通过多维标度法,根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的距离矩阵,并根据每个预先得到的质差用户的距离矩阵,绘制每个预先得到的质差用户的多维标度图;
展现绘制的多维标度图,并根据在操作界面接收到的修改指令,修改每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
其中,通过多维标度法,根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的距离矩阵的步骤,包括:
通过公式H=(hij)计算得到预先得到的质差用户的距离矩阵;其中,H表示预先得到的质差用户的距离矩阵。
其中,从每个预先得到的质差用户的每个聚类的指标变量所包含的第二数值型指标中,筛选出每个预先得到的质差用户的每个聚类的代表指标的步骤,包括:
给每个预先得到的质差用户的每个聚类的指标变量建立因子分析模型;
通过因子分析方法,根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型,得到每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵;
根据预设的特征根的累计方差贡献率的阈值,以及每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵,确定出每个预先得到的质差用户的每个聚类的指标变量的公共因子数量;
对每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵
进行正交旋转,并根据每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,计算出每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度;
根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度,以及每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,给每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型;
根据每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型,从每个预先得到的质差用户的每个聚类的指标变量包含的第二数值型指标中筛选出相关性贡献度最高的第二数值型指标,并将该第二数值型指标作为该聚类的代表指标。
其中,根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度,以及每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,给每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型的步骤,包括:
通过公式得到每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型;其中,RCtq表示第t个聚类的第q个第二数值型指标的相关性贡献度模型,Lt表示聚类的指标变量中第二数值型指标的数量,u表示聚类的指标变量的公共因子数量,b表示公共因子的序号,表示聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,Conbt表示聚类的指标变量的第b个公共因子的方差贡献度,T表示聚类数量;
其中,根据筛选出的代表指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤,包括:
根据每个预先得到的质差用户的每个聚类的代表指标,统计每个第二
数值型指标被筛选为代表指标的次数;
根据统计得到的次数,按照次数从高至低的顺序,对被筛选为代表指标的第二数值型指标进行排序;
根据在操作界面接收到的选择指令,从被筛选为代表指标的第二数值型指标中筛选出质差记录模型中的质差指标;
根据筛选出的质差指标,获取质差记录模型中质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,根据筛选出的质差指标,获取质差记录模型中质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤,包括:
将质差记录模型中质差指标的阈值设为第一预设值,并根据筛选出的质差指标以及设为第一预设值的质差指标的阈值构成的质差记录模型,确定每个预先得到的质差用户的每条观看记录的质差记录分布的标记;
根据每个预先得到的质差用户的每条观看记录的质差记录分布的标记,统计每个预先得到的质差用户的质差记录数量,并根据每个预先得到的质差用户的质差记录数量,计算每个预先得到的质差用户的质差记录占每个预先得到的质差用户的观看记录的数量的比值;
将质差用户模型中的质差记录占观看记录的数量的比重的阈值设为第二预设值,并确定每个预先得到的质差用户的比值大于或等于第二预设值;
控制质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型从多个第一用户和/或多个第二用户中筛选出质差用户;其中,第一用户的类型为无质差用户,第二用户的类型为质差用户;
获取质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型筛选质差用户的准确率,若准确率达到第三预设值,则将第一预设值作为质差指标
的阈值,并将第二预设值作为质差记录占观看记录的数量的比重的阈值;
若准确率未达到第三预设值,则调整第二预设值的大小,并根据调整后的第二预设值,调整第一预设值,直至质差指标的阈值为调整后的第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为调整后的第二预设值的质差用户模型筛选质差用户准确率达到第三预设值,并将调整后的第一预设值作为质差指标的阈值,以及将调整后的第二预设值作为质差记录占观看记录的数量的比重的阈值。
其中,根据筛选出的质差指标以及设为第一预设值的质差指标的阈值构成的质差记录模型,确定每个预先得到的质差用户的每条观看记录的质差记录分布的标记的步骤,包括:
若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值匹配,则将该观看记录的质差记录分布的标记设为1;
若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值不匹配,则将该观看记录的质差记录分布的标记设为0。
其中,确定每个预先得到的质差用户的比值大于或等于第二预设值的步骤,包括:
判断预先得到的质差用户的比值是否大于或等于第二预设值;
若预先得到的质差用户的比值小于第二预设值,则调整第一预设值,直至预先得到的质差用户的比值大于或等于第二预设值。
本发明的实施例还提供了一种监测交互式网络电视IPTV用户状态的装置,包括:
第一获取模块,设置为获取各IPTV用户的节目观看数据;其中,节目观看数据包括多条观看记录,每条观看记录包括多个第一数值型指标;
筛选模块,设置为根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录;
确定模块,设置为根据每个IPTV用户的多条观看记录中的质差记录,
确定出每个IPTV用户的状态。
在本发明实施例中,还提供了一种计算机存储介质,该计算机存储介质可以存储有执行指令,该执行指令用于执行上述实施例中的监测交互式网络电视IPTV用户状态方法的实现。
本发明实施例的上述方案至少包括以下有益效果:
在本发明的实施例中,通过根据各IPTV用户的多条观看记录中的第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录,并根据筛选出的质差记录,确定出每个IPTV用户的状态,解决了运营商无法及时、准确的确定出IPTV***中的各用户的状态,使得在IPTV***中出现感知恶化的用户时,难以及时对IPTV***进行网络优化,影响用户体验的问题,达到了使运营商及时、准确的确定出IPTV***中的各用户的状态,且在IPTV***中出现感知恶化的用户时,能及时对IPTV***进行网络优化,提升用户体验的效果。
图1为本发明第一实施例中监测交互式网络电视IPTV用户状态的方法的流程图;
图2为本发明第一实施例中获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的流程图;
图3为本发明第一实施例中的具体实例中对质差用户进行精细化分析的流程图;
图4为本发明第一实施例中的具体实例中绘制的聚类结果谱系图;
图5为本发明第一实施例中的具体实例中绘制的多维标度图;
图6为本发明第二实施例中监测交互式网络电视IPTV用户状态的装置的结构示意图;
图7为本发明第三实施例中IPTV数据分析架构的示意图。
下面将参照附图更详细地描述本公开的示例性实施例。虽然附图中显示了本公开的示例性实施例,然而应当理解,可以以各种形式实现本公开而不应被这里阐述的实施例所限制。相反,提供这些实施例是为了能够更透彻地理解本公开,并且能够将本公开的范围完整的传达给本领域的技术人员。
第一实施例
如图1所示,本发明的第一实施例提供了一种监测交互式网络电视IPTV用户状态的方法,该方法包括:
步骤101,获取各IPTV用户的节目观看数据。
其中,节目观看数据包括多条观看记录,每条观看记录包括多个第一数值型指标,且各第一数值型指标互不相同。具体的,该第一数值型指标可以为码率(avg-bit-rate)、组播缓冲下溢次数(multi-abend-numbers)、延时(df)、单播缓冲下溢次数(vod-abend-numbers)、丢包因子(mlr)、请求次数(req-numbers)、抖动(jitter)、切换时间(acc-avr-time)、可用率(can-use-rate)、单播申请失败总次数(vod-fail-numbers)、下溢次数(abend-numbers)、总的播放错误次数(play-error-numbers)、播放时长(play-time)或上溢次数(overflow-numbers)等。
步骤102,根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录。
其中,上述质差记录是指用户在观看节目的过程中,感觉节目质量恶化(例如卡顿花屏等)时的观看记录。
在本发明的第一实施例中,可通过检测IPTV用户的每条观看记录中的多个第一数值型指标的数值,是否与质差记录模型中的质差指标的阈值匹配的方式,筛选出每个IPTV用户的多条观看记录中的质差记录。具体的,若检测出IPTV用户的观看记录中的多个第一数值型指标的数值与质
差记录模型中的质差指标的阈值匹配,则确定该观看记录为质差记录。其中,质差记录模型为f1=F(Q′1>φ1,Q′2>φ2,...,Q′i>φi),i=Q′,其中,f1表示质差记录模型,Q′1表示质差记录模型中的第一个质差指标,φ1表示第一个质差指标的阈值,Q′2表示质差记录模型中的第二个质差指标,φ2表示第二个质差指标的阈值,Q′i表示质差记录模型中的第i个质差指标,φi表示第i个质差指标的阈值,Q′表示质差记录模型中质差指标的数量,F是指各质差指标进行组合排列的关系。
需要说明的是,质差记录模型中的各质差指标互不相同,且与第一数值型指标类似,可以为码率(avg-bit-rate)、组播缓冲下溢次数(multi-abend-numbers)、延时(df)、单播缓冲下溢次数(vod-abend-numbers)、丢包因子(mlr)、请求次数(req-numbers)、抖动(jitter)、切换时间(acc-avr-time)、可用率(can-use-rate)、单播申请失败总次数(vod-fail-numbers)、下溢次数(abend-numbers)、总的播放错误次数(play-error-numbers)、播放时长(play-time)或上溢次数(overflow-numbers)等。在此,以一具体例子进一步阐述上述步骤102,例如,质差记录模型中包括两个质差指标,分别为抖动(用Q′1表示)和延时(用Q′2表示),质差记录模型Q′1>φ1表示抖动大于4,Q′2>φ2表示延时大于10,此时,若IPTV用户的观看记录中的多个第一数值型指标中的抖动的数值为5和延时的数值为12,则认为该条观看记录为质差记录。
步骤103,根据每个IPTV用户的多条观看记录中的质差记录,确定出每个IPTV用户的状态。
在本发明的第一实施例中,步骤103的具体实现方式为:首先根据IPTV用户的多条观看记录中的质差记录,确定IPTV用户的每条观看记录的质差记录分布的标记,具体可通过公式确定IPTV用户的每条观看记录的质差记录分布的标记;其中,di表示IPTV用户的第i条观看记录的质差记录分布的标记,Di=f1表示第i条观看记录的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值匹配,Di≠f1表示第i条观看记录的多个第一数值型指标的数值与质差记录模型中的质差指
标的阈值不匹配;然后根据IPTV用户的每条观看记录的质差记录分布的标记,通过质差用户模型确定出IPTV用户的f2的值;其中,f2表示质差用户模型,D表示IPTV用户的观看记录的数量,表示质差记录占观看记录的数量的比重的阈值,其中,若f2的值为1,则确定该IPTV用户为质差用户。而若f2的值为0,则确定该IPTV用户为非质差用户。
其中,质差用户是指在观看节目的过程中,感觉节目质量恶化(例如卡顿花屏等)的IPTV用户,相应的,非质差用户是指在观看节目的过程中,感觉节目质量优良的IPTV用户。
需要说明的是,在本发明的第一实施例中,确定出的质差用户可以为卡顿花屏的质差用户,当然也可以是其他类型的质差用户。其确定出的质差用户类型主要取决于质差记录模型中各个质差指标的类型。具体的,若质差记录模型中各个质差指标是用于评判卡顿花屏的质差用户的,那么确定出的质差用户类型便为卡顿花屏的质差用户,类似的,若质差记录模型中各个质差指标是用于评判其他类型的质差用户的,那么确定出的质差用户类型便为其他类型的质差用户。
在本发明的第一实施例中,通过根据各IPTV用户的多条观看记录中的第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录,并根据筛选出的质差记录,确定出每个IPTV用户的状态,解决了运营商无法及时、准确的确定出IPTV***中的各用户的状态,使得在IPTV***中出现感知恶化的用户时,难以及时对IPTV***进行网络优化,影响用户体验的问题,达到了使运营商及时、准确的确定出IPTV***中的各用户的状态,且在IPTV***中出现感知恶化的用户时,能及时对IPTV***进行网络优化,提升用户体验的效果。
其中,在本发明的第一实施例中,在步骤102之前,上述方法还包括:根据预先得到的多个质差用户的观看记录所包含的多个第二数值型指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤。
如图2所示,根据预先得到的多个质差用户的观看记录所包含的多个第二数值型指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤包
括:
步骤201,根据每个预先得到的质差用户的多个第二数值型指标,得到每个预先得到的质差用户的指标相关性矩阵。
其中,上述多个第二数值型指标互不相同,且与第一数值型指标类似,可以为码率(avg-bit-rate)、组播缓冲下溢次数(multi-abend-numbers)、延时(df)、单播缓冲下溢次数(vod-abend-numbers)、丢包因子(mlr)、请求次数(req-numbers)、抖动(jitter)、切换时间(acc-avr-time)、可用率(can-use-rate)、单播申请失败总次数(vod-fail-numbers)、下溢次数(abend-numbers)、总的播放错误次数(play-error-numbers)、播放时长(play-time)或上溢次数(overflow-numbers)等。
步骤202,根据每个预先得到的质差用户的指标相关性矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
步骤203,从每个预先得到的质差用户的每个聚类的指标变量所包含的第二数值型指标中,筛选出每个预先得到的质差用户的每个聚类的代表指标。
步骤204,根据筛选出的代表指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
可选的,在本发明的第一实施例中,上述步骤201的具体包括如下步骤:
第一步,对每个预先得到的质差用户的多个第二数值型指标进行标准化处理,得到每个预先得到的质差用户的多个标准化数值型指标。
具体的,可通过公式计算得到每个预先得到的质差用户的多个标准化数值型指标;其中,zdq表示第d条观看记录的第q个第二数值型指标的标准化数值型指标,zdq表示第d条观看记录的第q个第二数值型指标,表示
第q个第二数值型指标的样本均值,sq表示第q个第二数值型指标的样本标准差,Dn表示第n个预先得到的质差用户的观看记录的数量,Q表示第二数值型指标的维度。
第二步,计算每个预先得到的质差用户的每两个标准化数值型指标之间的相关性,得到每个预先得到的质差用户的指标相关性矩阵。
具体的,可通过公式计算得到每个预先得到的质差用户的每两个标准化数值型指标之间的相关性,紧接着通过公式R=(rij)计算得到每个预先得到的质差用户的指标相关性矩阵;其中,R表示预先得到的质差用户的指标相关性矩阵,rij表示第i个标准化数值型指标与第j个标准化数值型指标之间的相关性,zdi表示第d条观看记录的第i个第二数值型指标的标准化数值型指标,表示第i个第二数值型指标的样本均值,zdj表示第d条观看记录的第j个第二数值型指标的标准化数值型指标,表示第j个第二数值型指标的样本均值。
可选的,在本发明的第一实施例中,上述步骤202的具体包括如下步骤:
第一步,确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同。
其中,第一步的具体实现方式为:判断指标相关性矩阵中是否存在相同的多个行(即,判断是否存在重复行),并若指标相关性矩阵中存在相同的多个行,则根据在操作界面接收到的删除指令,将多个行删除至一个行,并删除被删除行对应的第二数值型指标,使指标相关性矩阵中各行互不相同。其中,删除指令可由管理员根据自己的经验在操作界面输入,其直接指示删除哪些行。
第二步,根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的每两个第二数值型指标之间的相似性距离,得到每个预先得到的质差用户的相似性距离矩阵。
其中,可通过公式sij=1-rij计算得到每两个第二数值型指标之间的相似性距离,紧接着通过公式计算得到预先得到的质差用户的相似性距离矩阵;其中,sij表示第i个第二数值型指标与第j个第二数值型指标之间的相似性距离,S表示预先得到的质差用户的相似性距离矩阵,Qn′表示第二数值型指的数量。
第三步,通过R型聚类法,根据每个预先得到的质差用户的相似性距离矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
其中,可具体通过R型离差平方和(Ward)聚类法,根据每个预先得到的质差用户的相似性距离矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
其中,Ward法包括类中离差平方和距离计算公式和类间离差平方和距离计算公式。其中,类中离差平方和距离计算公式为:类间离差平方和计算公式为:其中,Lt(t=1,...,T)表示第t个聚类指标变量Xt包含的第二数值型指标的数量,表示第t个聚类中的平均距离。该聚类的指标变量可表示为Xt表示第t个聚类的指标变量,T表示聚类数量,z1表示第t个聚类的指标变量中的第一个第二数值型指标,表示第t个聚类的指标变量中的第Lt个第二数值型指标。
需要说明的是,在确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标之后,上述方法还包括:根据每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标,绘制每个预先得到的质差用户的聚类结果谱系图,并展现绘制的聚类结果谱系图的步骤,从而便于管理员可快速、清楚的知道每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
需要进一步说明的是,在确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同之后,上述方法还包括如下步骤:
第一步,通过多维标度法,根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的距离矩阵,并根据每个预先得到的质差用户的距离矩阵,绘制每个预先得到的质差用户的多维标度图。
其中,可通过公式计算得到每两个第二数值型指标之间的距离,紧接着通过公式H=(hij)计算得到预先得到的质差用户的距离矩阵;其中,hij表示第i个第二数值型指标与第j个第二数值型指标之间的距离,H表示预先得到的质差用户的距离矩阵。需要说明的是,在绘制多维标度图时,管理员可根据需求设定维度,例如二维等。
第二步,展现绘制的多维标度图,并根据在操作界面接收到的修改指令,修改每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
在本发明的第一实施例中,若管理员根据多维标度图,发现之前确定出的每个(或某个)预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标不准确的话,管理员可通过在操作界面输入修改指令的方式,修改每个(或某个)预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
可选的,在本发明的第一实施例中,上述步骤203的具体包括如下步骤:
第一步,给每个预先得到的质差用户的每个聚类的指标变量建立因子分析模型。
第二步,通过因子分析方法,根据每个预先得到的质差用户的每个聚
类的指标变量的因子分析模型,得到每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵。
第三步,根据预设的特征根的累计方差贡献率的阈值,以及每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵,确定出每个预先得到的质差用户的每个聚类的指标变量的公共因子数量。
第四步,对每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵进行正交旋转,并根据每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,计算出每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度。
其中,首先可通过公式Λ2=Λ1
(u)T对初等载荷矩阵进行正交旋转,Λ2表示旋转后的初等载荷矩阵,其中Λ1
(u)为Λ1前u列,T为正交矩阵,旋转后的载荷因子为然后通过公式计算得到因子分析模型中各公共因子的方差贡献度,其中,Conbt表示第b个公共因子的方差贡献度。
第五步,根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度,以及每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,给每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型。
其中,可通过公式得到每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型,紧接着通过公式计算得到预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型,其中,RCtq表示第t个聚类的第q个
第二数值型指标的相关性贡献度模型,Lt表示聚类的指标变量中第二数值型指标的数量,u表示聚类的指标变量的公共因子数量,b表示公共因子的序号,表示聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,Conbt表示聚类的指标变量的第b个公共因子的方差贡献度,T表示聚类数量,RCt表示聚类的指标变量建立相关性贡献度模型。
第六步,根据每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型,从每个预先得到的质差用户的每个聚类的指标变量包含的第二数值型指标中筛选出相关性贡献度最高的第二数值型指标,并将该第二数值型指标作为该聚类的代表指标。
可选的,在本发明的第一实施例中,上述步骤204的具体包括如下步骤:
第一步,根据每个预先得到的质差用户的每个聚类的代表指标,统计每个第二数值型指标被筛选为代表指标的次数。
第二步,根据统计得到的次数,按照次数从高至低的顺序,对被筛选为代表指标的第二数值型指标进行排序。
第三步,根据在操作界面接收到的选择指令,从被筛选为代表指标的第二数值型指标中筛选出质差记录模型中的质差指标。
其中,管理员可根据自己的经验输入选择指令,该选择指令指示选择能够反映网络状况的质差指标。
第四步,根据筛选出的质差指标,获取质差记录模型中质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,第四步的具体实现方式为:
首先,将质差记录模型中质差指标的阈值设为第一预设值(其中,若质差记录模型中包括多个质差指标,则每个质差指标都对应一个第一预设值,且各预设值可互不相同),并根据筛选出的质差指标以及设为第一预设值的质差指标的阈值构成的质差记录模型,确定每个预先得到的质差用户的每条观看记录的质差记录分布的标记。其中,若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设
值的质差指标的阈值匹配,则将该观看记录的质差记录分布的标记设为1;若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值不匹配,则将该观看记录的质差记录分布的标记设为0;
其次,根据每个预先得到的质差用户的每条观看记录的质差记录分布的标记,统计每个预先得到的质差用户的质差记录数量,并根据每个预先得到的质差用户的质差记录数量,计算每个预先得到的质差用户的质差记录占每个预先得到的质差用户的观看记录的数量的比值;
其次,将质差用户模型中的质差记录占观看记录的数量的比重的阈值设为第二预设值,并确定每个预先得到的质差用户的比值大于或等于第二预设值。其中,确定每个预先得到的质差用户的比值大于或等于第二预设值的步骤包括:判断预先得到的质差用户的比值是否大于或等于第二预设值;若预先得到的质差用户的比值小于第二预设值,则调整第一预设值,直至预先得到的质差用户的比值大于或等于第二预设值。
其次,控制质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型从多个第一用户和/或多个第二用户中筛选出质差用户;其中,第一用户的类型为无质差用户,第二用户的类型为质差用户;
其次,获取质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型筛选质差用户的准确率,若准确率达到第三预设值(例如70%),则将第一预设值作为质差指标的阈值,并将第二预设值作为质差记录占观看记录的数量的比重的阈值;
其次,若准确率未达到第三预设值,则调整第二预设值的大小,并根据调整后的第二预设值,调整第一预设值,直至质差指标的阈值为调整后的第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为调整后的第二预设值的质差用户模型筛选质差用户准确率达到
第三预设值,并将调整后的第一预设值作为质差指标的阈值,以及将调整后的第二预设值作为质差记录占观看记录的数量的比重的阈值。
在本发明的实施例中,以一具体实例进一步阐述上述获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤。
在该实例中,以卡顿花屏的质差用户为例,且以一个卡顿花屏的质差用户为例进行精细化分析,其中,该质差用户有32条观看记录,且该质差用户的第二数值型指标的维度为23。如图3所示,那么对该质差用户进行精细化分析的步骤如下:
步骤301,根据质差用户的多个第二数值型指标,得到指标相关性矩阵。
其中,上述多个第二数值型指标互不相同,且与第一数值型指标类似,可以为码率(avg-bit-rate)、组播缓冲下溢次数(multi-abend-numbers)、延时(df)、单播缓冲下溢次数(vod-abend-numbers)、丢包因子(mlr)、请求次数(req-numbers)、抖动(jitter)、切换时间(acc-avr-time)、可用率(can-use-rate)、单播申请失败总次数(vod-fail-numbers)、下溢次数(abend-numbers)、总的播放错误次数(play-error-numbers)、播放时长(play-time)或上溢次数(overflow-numbers)等。
步骤302,根据指标相关性矩阵和预设的聚类数量,确定出每个聚类的指标变量中包含的第二数值型指标。
步骤303,从每个聚类的指标变量所包含的第二数值型指标中,筛选出每个聚类的代表指标。
步骤304,根据筛选出的代表指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
可选的,在本发明的第一实施例中,上述步骤301的具体包括如下步骤:
第一步,对多个第二数值型指标进行标准化处理,得到多个标准化数
值型指标。
具体的,可通过公式计算得到质差用户的多个标准化数值型指标;其中,zdq表示第d条观看记录的第q个第二数值型指标的标准化数值型指标,zdq表示第d条观看记录的第q个第二数值型指标,表示第q个第二数值型指标的样本均值,sq表示第q个第二数值型指标的样本标准差,Dn表示质差用户的观看记录的数量,Q表示第二数值型指标的维度,Dn=32,Q=23。
第二步,计算每两个标准化数值型指标之间的相关性,得到指标相关性矩阵。
具体的,可通过公式计算得到质差用户的每两个标准化数值型指标之间的相关性,紧接着通过公式R=(rij)计算得到质差用户的指标相关性矩阵;其中,R表示质差用户的指标相关性矩阵,rij表示第i个标准化数值型指标与第j个标准化数值型指标之间的相关性,zdi表示第d条观看记录的第i个第二数值型指标的标准化数值型指标,表示第i个第二数值型指标的样本均值,zdi表示第d条观看记录的第j个第二数值型指标的标准化数值型指标,表示第j个第二数值型指标的样本均值。
可选的,在本发明的第一实施例中,上述步骤302的具体包括如下步骤:
第一步,确定质差用户的指标相关性矩阵中各行互不相同。
其中,第一步的具体实现方式为:判断指标相关性矩阵中是否存在相同的多个行(即,判断是否存在重复行),并若指标相关性矩阵中存在相同的多个行,则根据在操作界面接收到的删除指令,将多个行删除至一个行,并删除被删除行对应的第二数值型指标,使指标相关性矩阵中各行互
不相同。其中,删除指令可由管理员根据自己的经验在操作界面输入,其直接指示删除哪些行。
第二步,根据质差用户的指标相关性矩阵,计算质差用户的每两个第二数值型指标之间的相似性距离,得到质差用户的相似性距离矩阵。
其中,可通过公式sij=1-rij计算得到每两个第二数值型指标之间的相似性距离,紧接着通过公式计算得到质差用户的相似性距离矩阵;其中,sij表示第i个第二数值型指标与第j个第二数值型指标之间的相似性距离,S表示质差用户的相似性距离矩阵,Qn′表示第二数值型指的数量,需要说明的是,由于步骤301得到的指标相关性矩阵中存在重复行,删除了指标相关性矩阵中的重复行,并删除对应的第二数值型指标,因此Qn′的值与Q的值不相同,此处,Qn′=14。
第三步,通过R型聚类法,根据质差用户的相似性距离矩阵和预设的聚类数量,确定出质差用户的每个聚类的指标变量中包含的第二数值型指标。
其中,可具体通过R型离差平方和(Ward)聚类法,根据质差用户的相似性距离矩阵和预设的聚类数量(例如5),确定出质差用户的每个聚类的指标变量中包含的第二数值型指标。
其中,Ward法包括类中离差平方和距离计算公式和类间离差平方和距离计算公式。其中,类中离差平方和距离计算公式为:类间离差平方和计算公式为:其中,Lt(t=1,...,T)表示第t个聚类指标变量Xt包含的第二数值型指标的数量,表示第t个聚类中的平均距离。该类的指标变量可表示为Xt表示第t个聚类的指标变量,T表示聚类数量(此处,T=5),z1表示第t个聚类的指标变量中的第一个第二数值型指标,表示第t个聚类的指标变量中的第Lt个第二数值型指标。
需要说明的是,在确定出质差用户的每个聚类的指标变量中包含的第
二数值型指标之后,上述方法还包括:根据质差用户的每个聚类的指标变量中包含的第二数值型指标,绘制质差用户的聚类结果谱系图,并展现绘制的聚类结果谱系图的步骤,从而便于管理员可快速、清楚的知道质差用户的每个聚类的指标变量中包含的第二数值型指标。其中,绘制的聚类结果谱系图如图4所示,图4中横坐标轴中的1表示avg-bit-rate,2表示df,3表示multi-abend-numbers,4表示jitter,5表示can-use-rate,6表示abend-numbers,7表示play-time,8表示vod-abend-numbers,9表示req-numbers,10表示acc-avr-time,11表示vod-fail-numbers,12表示play-error-numbers,13表示mlr,14表示overflow-numbers。且从图4中可看出,abend-numbers、vod-abend-numbers、multi-abend-numbers可分为第一类;play-error-numbers、mlr可分为第二类;avg-bit-rate、df、jitter可分为第三类;overflow-numbers单独可分为第四类,req-numbers等剩余第二数值型指标可分为第五类。
需要进一步说明的是,在确定质差用户的指标相关性矩阵中各行互不相同之后,上述方法还包括如下步骤:
第一步,通过多维标度法,根据质差用户的指标相关性矩阵,计算质差用户的距离矩阵,并根据质差用户的距离矩阵,绘制质差用户的多维标度图。
其中,可通过公式计算得到每两个第二数值型指标之间的距离,紧接着通过公式H=(hij)计算得到质差用户的距离矩阵;其中,hij表示第i个第二数值型指标与第j个第二数值型指标之间的距离,H表示质差用户的距离矩阵。需要说明的是,在绘制多维标度图时,管理员可根据需求设定维度,例如二维等。其中,若将维度设为二维,绘制的多维标度图如图5所示。
第二步,展现绘制的多维标度图,并根据在操作界面接收到的修改指令,修改质差用户的每个聚类的指标变量中包含的第二数值型指标。
在本发明的第一实施例中,若管理员根据多维标度图,发现之前确定
出的质差用户的每个聚类的指标变量中包含的第二数值型指标不准确的话,管理员可通过在操作界面输入修改指令的方式,修改质差用户的每个聚类的指标变量中包含的第二数值型指标。
可选的,在本发明的第一实施例中,上述步骤303的具体包括如下步骤:
第一步,给质差用户的每个聚类的指标变量建立因子分析模型。
第二步,通过因子分析方法,根据质差用户的每个聚类的指标变量的因子分析模型,得到质差用户的每个聚类的指标变量的初等载荷矩阵。
第三步,根据预设的特征根的累计方差贡献率的阈值,以及质差用户的每个聚类的指标变量的初等载荷矩阵,确定出质差用户的每个聚类的指标变量的公共因子数量。
第四步,对质差用户的每个聚类的指标变量的初等载荷矩阵进行正交旋转,并根据质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,计算出质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度。
其中,首先可通过公式Λ2=Λ1
(u)T对初等载荷矩阵进行正交旋转,Λ2表示旋转后的初等载荷矩阵,其中Λ1
(u)为Λ1前u列,T为正交矩阵,旋转后的载荷因子为然后通过公式计算得到
因子分析模型中各公共因子的方差贡献度,其中,Conbt表示第b个公共因子的方差贡献度。
第五步,根据质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度,以及质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,给质差用户的每个聚类的指标变量建立相关性贡献度模型。
其中,可通过公式得到质差用户的每个聚类的指标变量建立相关性贡献度模型,紧接着通过公式计算得到质差用户的每个聚类的指标变量建立相关性贡献度模型,其中,RCtq表示第t个聚类的第q个第二数值型指标的相关性贡献度模型,Lt表示聚类的指标变量中第二数值型指标的数量,u表示聚类的指标变量的公共因子数量,b表示公共因子的序号,表示聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,Conbt表示聚类的指标变量的第b个公共因子的方差贡献度,T表示聚类数量,RCt表示聚类的指标变量建立相关性贡献度模型。
第六步,根据质差用户的每个聚类的指标变量建立相关性贡献度模型,从质差用户的每个聚类的指标变量包含的第二数值型指标中筛选出相关性贡献度最高的第二数值型指标,并将该第二数值型指标作为该聚类的代表指标。
可选的,在本发明的第一实施例中,上述步骤304的具体包括如下步骤:
第一步,根据质差用户的每个聚类的代表指标,统计每个第二数值型指标被筛选为代表指标的次数。
第二步,根据统计得到的次数,按照次数从高至低的顺序,对被筛选为代表指标的第二数值型指标进行排序。
第三步,根据在操作界面接收到的选择指令,从被筛选为代表指标的第二数值型指标中筛选出质差记录模型中的质差指标。
其中,管理员可根据自己的经验输入选择指令,该选择指令指示选择
能够反映网络状况的质差指标。
第四步,根据筛选出的质差指标,获取质差记录模型中质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,第四步的具体实现方式为:
首先,将质差记录模型中质差指标的阈值设为第一预设值(其中,若质差记录模型中包括多个质差指标,则每个质差指标都对应一个第一预设值,且各预设值可互不相同),并根据筛选出的质差指标以及设为第一预设值的质差指标的阈值构成的质差记录模型,确定质差用户的每条观看记录的质差记录分布的标记。其中,若质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值匹配,则将该观看记录的质差记录分布的标记设为1;若质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值不匹配,则将该观看记录的质差记录分布的标记设为0;
其次,根据质差用户的每条观看记录的质差记录分布的标记,统计质差用户的质差记录数量,并根据质差用户的质差记录数量,计算质差用户的质差记录占质差用户的观看记录的数量的比值;
其次,将质差用户模型中的质差记录占观看记录的数量的比重的阈值设为第二预设值,并确定质差用户的比值大于或等于第二预设值。其中,确定质差用户的比值大于或等于第二预设值的步骤包括:判断质差用户的比值是否大于或等于第二预设值;若质差用户的比值小于第二预设值,则调整第一预设值,直至质差用户的比值大于或等于第二预设值。
其次,控制质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型从多个第一用户和/或多个第二用户中筛选出质差用户;其中,第一用户的类型为无质差用户,第二用户的类型为质差用户;
其次,获取质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型筛
选质差用户的准确率,若准确率达到第三预设值(例如70%),则将第一预设值作为质差指标的阈值,并将第二预设值作为质差记录占观看记录的数量的比重的阈值;
其次,若准确率未达到第三预设值,则调整第二预设值的大小,并根据调整后的第二预设值,调整第一预设值,直至质差指标的阈值为调整后的第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为调整后的第二预设值的质差用户模型筛选质差用户准确率达到第三预设值,并将调整后的第一预设值作为质差指标的阈值,以及将调整后的第二预设值作为质差记录占观看记录的数量的比重的阈值。
需要说明的是,在本发明的第一实施例中,使用本发明的方法对5465个无质差用户进行判断筛选时,误判率仅为6.8%,但对59个卡顿花屏的质差用户的预测准确率高达66.1%。此外,本发明的方法不仅通过多维标度法、R型Ward聚类法实现对用户观看记录数据的高维指标进行联合聚类,还通过逆向因子分析法从指标中选举出代表指标以实现有效降维。统计代表指标并筛选出质差指标后,根据质差指标排列组合建立质差记录模型和质差用户模型,并不断迭代优化质差记录模型和质差用户模型的参数(即质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值),使其准确率与误判率满足要求。
第二实施例
如图6所示,本发明的第二实施例提供了一种监测交互式网络电视IPTV用户状态的装置,包括:
第一获取模块601,设置为获取各IPTV用户的节目观看数据;其中,节目观看数据包括多条观看记录,每条观看记录包括多个第一数值型指标;
筛选模块602,设置为根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录;
确定模块603,设置为根据每个IPTV用户的多条观看记录中的质差
记录,确定出每个IPTV用户的状态。
其中,筛选模块602包括:
第一筛选单元,设置为检测IPTV用户的每条观看记录中的多个第一数值型指标的数值,是否与质差记录模型中的质差指标的阈值匹配,并若IPTV用户的观看记录中的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值匹配,则触发第二筛选单元;
第二筛选单元,设置为根据第一筛选单元的触发,确定该观看记录为质差记录。
其中,质差记录模型为f1=F(Q′1>φ1,Q′2>φ2,...,Q′i>φi),i=Q′,其中,f1表示质差记录模型,Q′1表示质差记录模型中的第一个质差指标,φ1表示第一个质差指标的阈值,Q′2表示质差记录模型中的第二个质差指标,φ2表示第二个质差指标的阈值,Q′i表示质差记录模型中的第i个质差指标,φi表示第i个质差指标的阈值,Q′表示质差记录模型中质差指标的数量。
其中,确定模块603包括:
第三筛选单元,设置为根据IPTV用户的多条观看记录中的质差记录,确定IPTV用户的每条观看记录的质差记录分布的标记;
第四筛选单元,设置为根据IPTV用户的每条观看记录的质差记录分布的标记,通过质差用户模型确定出IPTV用户的f2的值;其中,f2表示质差用户模型,D表示IPTV用户的观看记录的数量,di表示IPTV用户的第i条观看记录的质差记录分布的标记,表示质差记录占观看记录的数量的比重的阈值;
第五筛选单元,设置为若f2的值为1,则确定该IPTV用户为质差用户;若f2的值为0,则确定该IPTV用户为非质差用户。
其中,
第三筛选单元,还设置为通过公式确定IPTV用户的每条观看记录的质差记录分布的标记;其中,Di=f1表示第i条观看记录的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值匹配,Di≠f1表示第i条观看记录的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值不匹配。
其中,装置还包括:
第二获取模块,设置为根据预先得到的多个质差用户的观看记录所包含的多个第二数值型指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,在本发明的第二实施例中,上述第二获取模块的功能可通过一IPTV分析***实现。
其中,第二获取模块包括:
第一获取单元,设置为根据每个预先得到的质差用户的多个第二数值型指标,得到每个预先得到的质差用户的指标相关性矩阵;
第二获取单元,设置为根据每个预先得到的质差用户的指标相关性矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标;
第三获取单元,设置为从每个预先得到的质差用户的每个聚类的指标变量所包含的第二数值型指标中,筛选出每个预先得到的质差用户的每个聚类的代表指标;
第四获取单元,设置为根据筛选出的代表指标,获取质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,第一获取单元包括:
第一获取子单元,设置为对每个预先得到的质差用户的多个第二数值型指标进行标准化处理,得到每个预先得到的质差用户的多个标准化数值型指标;
第二获取子单元,设置为计算每个预先得到的质差用户的每两个标准化数值型指标之间的相关性,得到每个预先得到的质差用户的指标相关性矩阵。
其中,
其中,
第二获取子单元,还设置为通过公式计算得到每个预先得到的质差用户的每两个标准化数值型指标之间的相关性;其中,rij表示第i个标准化数值型指标与第j个标准化数值型指标之间的相关性,zdi表示第d条观看记录的第i个第二数值型指标的标准化数值型指标,表示第i个第二数值型指标的样本均值,zdj表示第d条观看记录的第j个第二数值型指标的标准化数值型指标,表示第j个第二数值型指标的样本均值;
第二获取子单元,还设置为通过公式R=(rij)计算得到每个预先得到的质差用户的指标相关性矩阵;其中,R表示预先得到的质差用户的指标相关性矩阵。
其中,第二获取单元包括:
第三获取子单元,设置为确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同;
第四获取子单元,设置为根据每个预先得到的质差用户的指标相关性
矩阵,计算每个预先得到的质差用户的每两个第二数值型指标之间的相似性距离,得到每个预先得到的质差用户的相似性距离矩阵;
第五获取子单元,设置为通过R型聚类法,根据每个预先得到的质差用户的相似性距离矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
其中,
第四获取子单元,还设置为通过公式sij=1-rij计算得到每两个第二数值型指标之间的相似性距离;其中,sij表示第i个第二数值型指标与第j个第二数值型指标之间的相似性距离;
其中,装置还包括:
第一绘制模块,设置为根据每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标,绘制每个预先得到的质差用户的聚类结果谱系图,并展现绘制的聚类结果谱系图。
其中,
第三获取子单元,还设置为判断指标相关性矩阵中是否存在相同的多个行,并若指标相关性矩阵中存在相同的多个行,则根据在操作界面接收到的删除指令,将多个行删除至一个行,并删除被删除行对应的第二数值型指标,使指标相关性矩阵中各行互不相同。
其中,装置还包括:
第二绘制模块,设置为通过多维标度法,根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的距离矩阵,并根据每个预先得到的质差用户的距离矩阵,绘制每个预先得到的质差用户的多维标度图;
修改模块,设置为展现绘制的多维标度图,并根据在操作界面接收到的修改指令,修改每个预先得到的质差用户的每个聚类的指标变量中包含
的第二数值型指标。
其中,第二绘制模块包括:
第二绘制单元,设置为通过公式H=(hij)计算得到预先得到的质差用户的距离矩阵;其中,H表示预先得到的质差用户的距离矩阵。
其中,第三获取单元包括:
第六获取子单元,设置为给每个预先得到的质差用户的每个聚类的指标变量建立因子分析模型;
第七获取子单元,设置为通过因子分析方法,根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型,得到每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵;
第八获取子单元,设置为根据预设的特征根的累计方差贡献率的阈值,以及每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵,确定出每个预先得到的质差用户的每个聚类的指标变量的公共因子数量;
第九获取子单元,设置为对每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵进行正交旋转,并根据每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,计算出每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度;
第十获取子单元,设置为根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度,以及每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,给每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型;
第十一获取子单元,设置为根据每个预先得到的质差用户的每个聚类
的指标变量建立相关性贡献度模型,从每个预先得到的质差用户的每个聚类的指标变量包含的第二数值型指标中筛选出相关性贡献度最高的第二数值型指标,并将该第二数值型指标作为该聚类的代表指标。
其中,
第十获取子单元,还设置为通过公式得到每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型;其中,RCtq表示第t个聚类的第q个第二数值型指标的相关性贡献度模型,Lt表示聚类的指标变量中第二数值型指标的数量,u表示聚类的指标变量的公共因子数量,b表示公共因子的序号,表示聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,Conbt表示聚类的指标变量的第b个公共因子的方差贡献度,T表示聚类数量;
其中,第四获取单元包括:
第十二获取子单元,设置为根据每个预先得到的质差用户的每个聚类的代表指标,统计每个第二数值型指标被筛选为代表指标的次数;
第十三获取子单元,设置为根据统计得到的次数,按照次数从高至低的顺序,对被筛选为代表指标的第二数值型指标进行排序;
第十四获取子单元,设置为根据在操作界面接收到的选择指令,从被筛选为代表指标的第二数值型指标中筛选出质差记录模型中的质差指标;
第十五获取子单元,设置为根据筛选出的质差指标,获取质差记录模型中质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
其中,
第十五获取子单元,还设置为将质差记录模型中质差指标的阈值设为第一预设值,并根据筛选出的质差指标以及设为第一预设值的质差指标的阈值构成的质差记录模型,确定每个预先得到的质差用户的每条观看记录
的质差记录分布的标记;
第十五获取子单元,还设置为根据每个预先得到的质差用户的每条观看记录的质差记录分布的标记,统计每个预先得到的质差用户的质差记录数量,并根据每个预先得到的质差用户的质差记录数量,计算每个预先得到的质差用户的质差记录占每个预先得到的质差用户的观看记录的数量的比值;
第十五获取子单元,还设置为将质差用户模型中的质差记录占观看记录的数量的比重的阈值设为第二预设值,并确定每个预先得到的质差用户的比值大于或等于第二预设值;
第十五获取子单元,还设置为控制质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型从多个第一用户和/或多个第二用户中筛选出质差用户;其中,第一用户的类型为无质差用户,第二用户的类型为质差用户;
第十五获取子单元,还设置为获取质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型筛选质差用户的准确率,若准确率达到第三预设值,则将第一预设值作为质差指标的阈值,并将第二预设值作为质差记录占观看记录的数量的比重的阈值;
第十五获取子单元,还设置为若准确率未达到第三预设值,则调整第二预设值的大小,并根据调整后的第二预设值,调整第一预设值,直至质差指标的阈值为调整后的第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为调整后的第二预设值的质差用户模型筛选质差用户准确率达到第三预设值,并将调整后的第一预设值作为质差指标的阈值,以及将调整后的第二预设值作为质差记录占观看记录的数量的比重的阈值。
其中,
第十五获取子单元,还设置为若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的
阈值匹配,则将该观看记录的质差记录分布的标记设为1;
第十五获取子单元,还设置为若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值不匹配,则将该观看记录的质差记录分布的标记设为0。
其中,
第十五获取子单元,还设置为判断预先得到的质差用户的比值是否大于或等于第二预设值,并若预先得到的质差用户的比值小于第二预设值,则调整第一预设值,直至预先得到的质差用户的比值大于或等于第二预设值。
在本发明的第二实施例中,通过根据各IPTV用户的多条观看记录中的第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录,并根据筛选出的质差记录,确定出每个IPTV用户的状态,解决了运营商无法及时、准确的确定出IPTV***中的各用户的状态,使得在IPTV***中出现感知恶化的用户时,难以及时对IPTV***进行网络优化,影响用户体验的问题,达到了使运营商及时、准确的确定出IPTV***中的各用户的状态,且在IPTV***中出现感知恶化的用户时,能及时对IPTV***进行网络优化,提升用户体验的效果。
需要说明的是,本发明第二实施例提供的监测交互式网络电视IPTV用户状态的装置是应用上述监测交互式网络电视IPTV用户状态的方法的装置,即上述方法的所有实施例均适用于该装置,且均能达到相同或相似的有益效果。
第三实施例
如图7所示,本发明的第三实施例提供了一种IPTV数据分析架构,包括:数据获取模块701、探针模块702、IPTV服务质量保障***(IQAS)以及IPTV分析***704。其中,数据获取模块701设置为抓取质差用户观看节目时的网络包,其可通过一libpacp模块实现;探针模块702设置为解析所抓取的网络包,并上报给IQAS703,使得IPTV分析***704可
从IQAS703中获取质差用户的数据进行分析,确定出质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
以上是本发明的优选实施方式,应当指出,对于本技术领域的普通技术人员来说,在不脱离本发明原理的前提下,还可以作出若干改进和润饰,这些改进和润饰也应视为本发明的保护范围。
本发明实施例提供的上述技术方案,通过根据各IPTV用户的多条观看记录中的第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录,并根据筛选出的质差记录,确定出每个IPTV用户的状态,解决了运营商无法及时、准确的确定出IPTV***中的各用户的状态,使得在IPTV***中出现感知恶化的用户时,难以及时对IPTV***进行网络优化,影响用户体验的问题,达到了使运营商及时、准确的确定出IPTV***中的各用户的状态,且在IPTV***中出现感知恶化的用户时,能及时对IPTV***进行网络优化。
Claims (23)
- 一种监测交互式网络电视IPTV用户状态的方法,包括:获取各IPTV用户的节目观看数据;其中,所述节目观看数据包括多条观看记录,每条观看记录包括多个第一数值型指标;根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录;根据每个IPTV用户的多条观看记录中的质差记录,确定出每个IPTV用户的状态。
- 如权利要求1所述的方法,其中,所述根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录的步骤,包括:检测IPTV用户的每条观看记录中的多个第一数值型指标的数值,是否与质差记录模型中的质差指标的阈值匹配;若IPTV用户的观看记录中的多个第一数值型指标的数值与质差记录模型中的质差指标的阈值匹配,则确定该观看记录为质差记录。
- 如权利要求2所述的方法,其中,所述质差记录模型为f1=F(Q′1>φ1,Q′2>φ2,...,Q′i>φi),i=Q′,其中,f1表示质差记录模型,Q′1表示质差记录模型中的第一个质差指标,φ1表示第一个质差指标的阈值,Q′2表示质差记录模型中的第二个质差指标,φ2表示第二个质差指标的阈值,Q′i表示质差记录模型中的第i个质差指标,φi表示第i个质差指标的阈值,Q′表示质差记录模型中质差指标的数量。
- 如权利要求5所述的方法,其中,在所述根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录的步骤之前,所述方法还包括:根据预先得到的多个质差用户的观看记录所包含的多个第二数值型指标,获取所述质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
- 如权利要求5所述的方法,其中,所述根据预先得到的多个质差用户的观看记录所包含的多个第二数值型指标,获取所述质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤,包括:根据每个预先得到的质差用户的多个第二数值型指标,得到每个预先 得到的质差用户的指标相关性矩阵;根据每个预先得到的质差用户的指标相关性矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标;从每个预先得到的质差用户的每个聚类的指标变量所包含的第二数值型指标中,筛选出每个预先得到的质差用户的每个聚类的代表指标;根据筛选出的代表指标,获取所述质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
- 如权利要求7所述的方法,其中,所述根据每个预先得到的质差用户的多个第二数值型指标,得到每个预先得到的质差用户的指标相关性矩阵的步骤,包括:对每个预先得到的质差用户的多个第二数值型指标进行标准化处理,得到每个预先得到的质差用户的多个标准化数值型指标;计算每个预先得到的质差用户的每两个标准化数值型指标之间的相关性,得到每个预先得到的质差用户的指标相关性矩阵。
- 如权利要求9所述的方法,其中,所述计算每个预先得到的质差用户的每两个标准化数值型指标之间的相关性,得到每个预先得到的质差用户的指标相关性矩阵的步骤,包括:通过公式(i,j=1,...,Q)计算得到每个预先得到的质差用户的每两个标准化数值型指标之间的相关性;其中,rij表示第i个标准化数值型指标与第j个标准化数值型指标之间的相关性,zdi表示第d条观看记录的第i个第二数值型指标的标准化数值型指标,表示第i个第二数值型指标的样本均值,zdj表示第d条观看记录的第j个第二数值型指标的标准化数值型指标,表示第j个第二数值型指标的样本均值;通过公式R=(rij)计算得到每个预先得到的质差用户的指标相关性矩阵;其中,R表示预先得到的质差用户的指标相关性矩阵。
- 如权利要求10所述的方法,其中,所述根据每个预先得到的质差用户的指标相关性矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标的步骤,包括:确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同;根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的每两个第二数值型指标之间的相似性距离,得到每个预先得到的质差用户的相似性距离矩阵;通过R型聚类法,根据每个预先得到的质差用户的相似性距离矩阵和预设的聚类数量,确定出每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
- 如权利要求11所述的方法,其中,所述方法还包括:根据每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标,绘制每个预先得到的质差用户的聚类结果谱系图,并展现绘制的聚类结果谱系图。
- 如权利要求11所述的方法,其中,所述确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同的步骤,包括:判断所述指标相关性矩阵中是否存在相同的多个行;若所述指标相关性矩阵中存在相同的多个行,则根据在操作界面接收到的删除指令,将所述多个行删除至一个行,并删除被删除行对应的第二数值型指标,使指标相关性矩阵中各行互不相同。
- 如权利要求11所述的方法,其中,在所述确定每个预先得到的质差用户的指标相关性矩阵中各行互不相同的步骤之后,所述方法还包括:通过多维标度法,根据每个预先得到的质差用户的指标相关性矩阵,计算每个预先得到的质差用户的距离矩阵,并根据每个预先得到的质差用户的距离矩阵,绘制每个预先得到的质差用户的多维标度图;展现绘制的多维标度图,并根据在操作界面接收到的修改指令,修改每个预先得到的质差用户的每个聚类的指标变量中包含的第二数值型指标。
- 如权利要求7所述的方法,其中,所述从每个预先得到的质差用户的每个聚类的指标变量所包含的第二数值型指标中,筛选出每个预先得到的质差用户的每个聚类的代表指标的步骤,包括:给每个预先得到的质差用户的每个聚类的指标变量建立因子分析模型;通过因子分析方法,根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型,得到每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵;根据预设的特征根的累计方差贡献率的阈值,以及每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵,确定出每个预先得到的质差用户的每个聚类的指标变量的公共因子数量;对每个预先得到的质差用户的每个聚类的指标变量的初等载荷矩阵进行正交旋转,并根据每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,计算出每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度;根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度,以及每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,给每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型;根据每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型,从每个预先得到的质差用户的每个聚类的指标变量包含的第二数值型指标中筛选出相关性贡献度最高的第二数值型指标,并将该第二数值型指标作为该聚类的代表指标。
- 如权利要求17所述的方法,其中,所述根据每个预先得到的质差用户的每个聚类的指标变量的因子分析模型中各公共因子的方差贡献度,以及每个预先得到的质差用户的每个聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,给每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型的步骤,包括:通过公式q=1,...,Lt,t=1,...,T得到每个预先得到的质差用户的每个聚类的指标变量建立相关性贡献度模型;其中,RCtq表示第t个聚类的第q个第二数值型指标的相关性贡献度模型,Lt表示聚类的指标变量中第二数值型指标的数量,u表示聚类的指标变量的公共因子数量,b表示公共因子的序号,表示聚类的指标变量的旋转后的初等载荷矩阵中的载荷因子,Conbt表示聚类的指标变量的第b个公共因子的方差贡献度,T表示聚类数量;
- 如权利要求7所述的方法,其中,所述根据筛选出的代表指标,获取所述质差记录模型中的质差指标和质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤,包括:根据每个预先得到的质差用户的每个聚类的代表指标,统计每个第二数值型指标被筛选为代表指标的次数;根据统计得到的次数,按照次数从高至低的顺序,对被筛选为代表指标的第二数值型指标进行排序;根据在操作界面接收到的选择指令,从被筛选为代表指标的第二数值型指标中筛选出所述质差记录模型中的质差指标;根据筛选出的质差指标,获取所述质差记录模型中质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值。
- 如权利要求19所述的方法,其中,所述根据筛选出的质差指标,获取所述质差记录模型中质差指标的阈值,以及质差用户模型中的质差记录占观看记录的数量的比重的阈值的步骤,包括:将所述质差记录模型中质差指标的阈值设为第一预设值,并根据筛选出的质差指标以及设为第一预设值的质差指标的阈值构成的质差记录模型,确定每个预先得到的质差用户的每条观看记录的质差记录分布的标记;根据每个预先得到的质差用户的每条观看记录的质差记录分布的标记,统计每个预先得到的质差用户的质差记录数量,并根据每个预先得到的质差用户的质差记录数量,计算每个预先得到的质差用户的质差记录占每个预先得到的质差用户的观看记录的数量的比值;将所述质差用户模型中的质差记录占观看记录的数量的比重的阈值设为第二预设值,并确定每个预先得到的质差用户的比值大于或等于所述第二预设值;控制质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型从多个第一用户和/或多个第二用户中筛选出质差用户;其中,所述第一用户的类型为无质差用户,所述第二用户的类型为质差用户;获取质差指标的阈值为第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为第二预设值的质差用户模型筛选质差用户的准确率,若所述准确率达到第三预设值,则将所述第一预设值作为质差指标的阈值,并将第二预设值作为质差记录占观看记录的数量的比重的阈值;若所述准确率未达到所述第三预设值,则调整所述第二预设值的大小,并根据调整后的第二预设值,调整所述第一预设值,直至质差指标的阈值为调整后的第一预设值的质差记录模型,以及质差记录占观看记录的数量的比重的阈值设为调整后的第二预设值的质差用户模型筛选质差用户准确率达到第三预设值,并将调整后的第一预设值作为质差指标的阈值,以及将调整后的第二预设值作为质差记录占观看记录的数量的比重的阈值。
- 如权利要求20所述的方法,其中,所述根据筛选出的质差指标以及设为第一预设值的质差指标的阈值构成的质差记录模型,确定每个预先得到的质差用户的每条观看记录的质差记录分布的标记的步骤,包括:若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值匹配,则将该观看记录的质差记录分布的标记设为1;若预先得到的质差用户的观看记录中的第二数值型指标的数值,与质差记录模型中设为第一预设值的质差指标的阈值不匹配,则将该观看记录的质差记录分布的标记设为0。
- 如权利要求21所述的方法,其中,所述确定每个预先得到的质差用户的比值大于或等于所述第二预设值的步骤,包括:判断预先得到的质差用户的比值是否大于或等于所述第二预设值;若预先得到的质差用户的比值小于所述第二预设值,则调整所述第一预设值,直至预先得到的质差用户的比值大于或等于所述第二预设值。
- 一种监测交互式网络电视IPTV用户状态的装置,包括:第一获取模块,设置为获取各IPTV用户的节目观看数据;其中,所述节目观看数据包括多条观看记录,每条观看记录包括多个第一数值型指标;筛选模块,设置为根据每个IPTV用户的每条观看记录中的多个第一数值型指标,筛选出每个IPTV用户的多条观看记录中的质差记录;确定模块,设置为根据每个IPTV用户的多条观看记录中的质差记录,确定出每个IPTV用户的状态。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610589120.9A CN107659845B (zh) | 2016-07-25 | 2016-07-25 | 一种监测交互式网络电视iptv用户状态的方法及装置 |
CN201610589120.9 | 2016-07-25 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2018019210A1 true WO2018019210A1 (zh) | 2018-02-01 |
Family
ID=61015734
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2017/094151 WO2018019210A1 (zh) | 2016-07-25 | 2017-07-24 | 一种监测交互式网络电视iptv用户状态的方法及装置 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN107659845B (zh) |
WO (1) | WO2018019210A1 (zh) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20200099899A (ko) * | 2019-02-15 | 2020-08-25 | 영남대학교 산학협력단 | 발열 소자, 이를 포함하는 아토마이저 및 이를 포함하는 전자식 니코틴 전달 시스템 |
CN113453076A (zh) * | 2020-03-24 | 2021-09-28 | ***通信集团河北有限公司 | 用户视频业务质量评估方法、装置、计算设备和存储介质 |
CN113691406A (zh) * | 2021-08-27 | 2021-11-23 | 中国电信股份有限公司 | 网络质量优化方法及装置、存储介质和电子设备 |
WO2024139573A1 (zh) * | 2022-12-27 | 2024-07-04 | 中兴通讯股份有限公司 | 阈值确定方法、装置、存储介质及电子装置 |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108984369A (zh) * | 2018-07-13 | 2018-12-11 | 厦门美图移动科技有限公司 | 卡顿预测方法、装置及移动终端 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101083063B1 (ko) * | 2010-09-09 | 2011-11-16 | 주식회사 케이티 | 비디오 체감 품질을 측정하는 방법 및 장치 |
CN102291617A (zh) * | 2011-09-03 | 2011-12-21 | 四川公用信息产业有限责任公司 | Iptv业务端到端故障诊断与定位平台 |
CN102972042A (zh) * | 2010-07-05 | 2013-03-13 | 三菱电机株式会社 | 影像质量管理*** |
CN104540018A (zh) * | 2014-12-17 | 2015-04-22 | 北京国双科技有限公司 | 网络电视视频异常播放数据的处理方法和装置 |
CN105635722A (zh) * | 2014-10-27 | 2016-06-01 | 青岛金讯网络工程有限公司 | 基于媒体丢包率指标的iptv业务健康度评价方法 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
IL210169A0 (en) * | 2010-12-22 | 2011-03-31 | Yehuda Binder | System and method for routing-based internet security |
CN102143388B (zh) * | 2011-04-22 | 2013-12-18 | 赛特斯信息科技股份有限公司 | Iptv用户体验质量评估装置及评估方法 |
-
2016
- 2016-07-25 CN CN201610589120.9A patent/CN107659845B/zh active Active
-
2017
- 2017-07-24 WO PCT/CN2017/094151 patent/WO2018019210A1/zh active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102972042A (zh) * | 2010-07-05 | 2013-03-13 | 三菱电机株式会社 | 影像质量管理*** |
KR101083063B1 (ko) * | 2010-09-09 | 2011-11-16 | 주식회사 케이티 | 비디오 체감 품질을 측정하는 방법 및 장치 |
CN102291617A (zh) * | 2011-09-03 | 2011-12-21 | 四川公用信息产业有限责任公司 | Iptv业务端到端故障诊断与定位平台 |
CN105635722A (zh) * | 2014-10-27 | 2016-06-01 | 青岛金讯网络工程有限公司 | 基于媒体丢包率指标的iptv业务健康度评价方法 |
CN104540018A (zh) * | 2014-12-17 | 2015-04-22 | 北京国双科技有限公司 | 网络电视视频异常播放数据的处理方法和装置 |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20200099899A (ko) * | 2019-02-15 | 2020-08-25 | 영남대학교 산학협력단 | 발열 소자, 이를 포함하는 아토마이저 및 이를 포함하는 전자식 니코틴 전달 시스템 |
KR102223201B1 (ko) * | 2019-02-15 | 2021-03-05 | 영남대학교 산학협력단 | 발열 소자, 이를 포함하는 아토마이저 및 이를 포함하는 전자식 니코틴 전달 시스템 |
CN113453076A (zh) * | 2020-03-24 | 2021-09-28 | ***通信集团河北有限公司 | 用户视频业务质量评估方法、装置、计算设备和存储介质 |
CN113691406A (zh) * | 2021-08-27 | 2021-11-23 | 中国电信股份有限公司 | 网络质量优化方法及装置、存储介质和电子设备 |
CN113691406B (zh) * | 2021-08-27 | 2022-09-02 | 中国电信股份有限公司 | 网络质量优化方法及装置、存储介质和电子设备 |
WO2024139573A1 (zh) * | 2022-12-27 | 2024-07-04 | 中兴通讯股份有限公司 | 阈值确定方法、装置、存储介质及电子装置 |
Also Published As
Publication number | Publication date |
---|---|
CN107659845B (zh) | 2021-08-03 |
CN107659845A (zh) | 2018-02-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2018019210A1 (zh) | 一种监测交互式网络电视iptv用户状态的方法及装置 | |
US20230057697A1 (en) | Detecting user interest in presented media items by observing volume change events | |
US11683234B2 (en) | Server outlier detection | |
US20190297379A1 (en) | Method and apparatus for enabling a loudness controller to adjust a loudness level of a secondary media data portion in a media content to a different loudness level | |
CN110909205B (zh) | 一种视频封面确定方法、装置、电子设备及可读存储介质 | |
CN108874832B (zh) | 目标评论确定方法及装置 | |
CN109753601B (zh) | 推荐信息点击率确定方法、装置及电子设备 | |
CN106992974B (zh) | 一种直播视频信息监控方法、装置和设备 | |
CN105163142B (zh) | 一种用户偏好确定方法、视频推荐方法和*** | |
US9760780B2 (en) | Thumbnail management | |
WO1998011494A1 (en) | Data correlation and analysis tool | |
CN110769283A (zh) | 一种视频推送方法、装置、计算机设备和存储介质 | |
CN105843876B (zh) | 多媒体资源的质量评估方法和装置 | |
CN111225246B (zh) | 一种视频推荐方法、装置及电子设备 | |
EP3070959A1 (en) | Methods and systems for content presentation optimization | |
JP2017010277A (ja) | 作業分析システム及び作業分析方法 | |
JP7502570B2 (ja) | 酒製品位置決め方法、酒製品情報管理方法及びその装置、デバイス及び記憶媒体 | |
CN113158048A (zh) | 一种基于特征识别的移动互联网浏览内容智能推荐方法、***、设备及存储介质 | |
US10630633B2 (en) | Real-time collection, analysis and presentation of aggregated selection data by users of social media | |
US11016467B2 (en) | Method and system for sensing fine changes in processing/equipment measurement data | |
CN113886272A (zh) | 一种应用程序的性能测试方法、***、装置和测试设备 | |
CN115909166A (zh) | 一种视频评估方法、装置、电子设备和存储介质 | |
CN110309361B (zh) | 一种视频评分的确定方法、推荐方法、装置及电子设备 | |
CN112906805A (zh) | 图像训练样本筛选和任务模型训练方法、装置及电子设备 | |
KR102690528B1 (ko) | 미디어 시청 데이터를 사용하여 디바이스 맵의 정확도를 향상시키는 시스템 및 방법 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 17833513 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 17833513 Country of ref document: EP Kind code of ref document: A1 |