CN106126544A - The put-on method of a kind of internet content and device - Google Patents
The put-on method of a kind of internet content and device Download PDFInfo
- Publication number
- CN106126544A CN106126544A CN201610424892.7A CN201610424892A CN106126544A CN 106126544 A CN106126544 A CN 106126544A CN 201610424892 A CN201610424892 A CN 201610424892A CN 106126544 A CN106126544 A CN 106126544A
- Authority
- CN
- China
- Prior art keywords
- access
- record
- characteristic
- access record
- example sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0255—Targeted advertisements based on user history
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0277—Online advertisement
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Theoretical Computer Science (AREA)
- Finance (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Economics (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Information Transfer Between Computers (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The put-on method of a kind of internet content and device, described method includes: obtain active user's history access record data in current site, and be multiple continuous access records by described history access record data cutting, each continuous access record includes at least one access record that the connected reference behavior of current site is formed by active user, and every accesses record and includes active user's access behavioral data to object in current site；Positive example sample and negative example sample is chosen from the access record that multiple continuous access records comprise；According to positive example sample and negative example sample, access characteristic from described history access record Data induction；Model training is carried out, to obtain multiple regression model based on accessing characteristic；Renewals based on multiple objects to be predicted access characteristic and the plurality of regression model, determine the input priority of the content recommendation associated with the plurality of object to be predicted.Such scheme makes the input of internet content more accurate.
Description
Technical field
The present invention relates to Internet technical field, particularly relate to put-on method and the device of a kind of internet content.
Background technology
Nowadays, along with the development of the Internet especially mobile Internet, the Internet can be supplied to the information of user more
Coming the abundantest, user can be concerned about different classes of information content by the Internet.Such as, for news content, Yong Huke
With by internet access sport category information, science and technology category information, finance and economic information and amusement category information etc..
The behavior on website in history according to user, such as browses, searches for, collection etc., can push it once to user
Object through accessing, such as the relevant content of commodity etc..In the prior art, it is recommended that method often join according to business and experience
Putting and simply process rule, such as processing the feature selected by rule is the time sequencing that user browses object, according to from currently
Time, by closely to remote order arrangement content recommendation release sequence, is estimated user and clicks on the probability of the object that the last time browses
Bigger.
But, not accurate enough to the preferential click order of content recommendation according to the user that the method for prior art is predicted, change
Sentence is talked about, and the content of relatively preferential recommendation can not represent the content that user priority is clicked on exactly.
Summary of the invention
Present invention solves the technical problem that put-on method and the device being to provide a kind of internet content, improve in the Internet
Hold the accuracy rate thrown in.
For solving above-mentioned technical problem, the embodiment of the present invention provides the put-on method of internet content, and described method includes:
Obtain active user's history access record data in current site, and by described history access record data cutting
For multiple continuous access records, each described continuous access record includes the described active user continuous visit to described current site
At least one access record that behavior of asking is formed, every described access record includes that described active user is in described current site
Access behavioral data to object；Positive example sample and negative example is chosen from the access record that the plurality of continuous access record comprises
Sample；According to described positive example sample and described negative example sample, access characteristic from described history access record Data induction；Base
Model training is carried out, to obtain multiple regression model in described access characteristic；Renewals based on multiple objects to be predicted are visited
Ask characteristic and the plurality of regression model, determine that the input of the content recommendation associated with the plurality of object to be predicted is excellent
First order.
Alternatively, from the access record that the plurality of continuous access record comprises, positive example sample and negative example sample packages are chosen
Include:
For the access record in each continuous access record, if the object of described access recording needle pair holding previously
Continuous access in record is accessed, then be described positive example sample by described access recording mark；
For the access record in each continuous access record, if the object of described access recording needle pair holding previously
Continuous access in record is accessed, but the most accessed in existing continuous accesses record, then by described access recording mark be
Described negative example sample.
Alternatively, described regression model is GBDT tree-model.
Alternatively, renewals based on multiple objects to be predicted access characteristic and the plurality of regression model, determine
With the input priority of the content recommendation that the plurality of object to be predicted associates, including:
Renewal based on object to be predicted accesses characteristic and the plurality of regression model obtains described object to be predicted
Score；
It is ranked up by from high to low according to described score, and determines and the plurality of object to be predicted according to described sequence
The input priority of the content recommendation of association.
Alternatively, described access characteristic includes that active user accesses the characteristic of described current site, object quilt
The characteristic accessed and described active user access the characteristic of described object.
The embodiment of the present invention also provides for the delivery device of a kind of internet content, and described device includes:
Acquiring unit, is suitable to the history access record data obtaining active user in current site, and described history is visited
Asking that record data cutting is multiple continuous access records, each described continuous access record includes that described active user works as described
At least one access record that the connected reference behavior of front website is formed, every described access record includes that described active user exists
Access behavioral data to object in described current site；
Choose unit, be suitable to choose positive example sample and negative example from the access record that the plurality of continuous access record comprises
Sample；
Conclude unit, be suitable to, according to described positive example sample and described negative example sample, return from described history access record data
Receive access characteristic；
Model training unit, is suitable to carry out model training based on described access characteristic, to obtain multiple regression model；
Determine unit, be suitable to renewal based on multiple objects to be predicted and access characteristic and the plurality of recurrence mould
Type, determines the input priority of the content recommendation associated with the plurality of object to be predicted.
Alternatively, choose unit described in include:
First labelling subelement, is suitable to for the access record in each continuous access record, if described access record
For object continuous access record previously in be accessed, then be described positive example sample by described access recording mark；
Second labelling subelement, is suitable to for the access record in each continuous access record, if described access record
For object continuous access record previously in be accessed, but the most accessed, then in existing continuous accesses record
It is described negative example sample by described access recording mark.
Alternatively, described regression model is GBDT model.
Alternatively, described determine that unit includes:
Score obtains subelement, is suitable to renewal based on object to be predicted and accesses characteristic and the plurality of regression model
Obtain the score of described object to be predicted；
Release sequence determines subelement, is suitable to be ranked up by from high to low according to described score, and according to described sequence
Determine the input priority of the content recommendation associated with the plurality of object to be predicted.
Alternatively, described access characteristic includes that active user accesses the characteristic of described current site, object quilt
The characteristic accessed and described active user access the characteristic of described object.
Compared with prior art, the technical scheme of the embodiment of the present invention has the advantages that
The technical scheme of the embodiment of the present invention by obtaining active user in the history access record data of current site, and
It is multiple continuous access records by described history access record data cutting, the access comprised from the plurality of continuous access record
Record is chosen positive example sample and negative example sample, each positive example sample and negative example sample is concluded and accesses characteristic, based on
The described access characteristic being marked as positive example or negative example carries out model training, to obtain multiple regression model, then passes through back
Updating of model and object to be predicted is returned to access characteristic and determine the content recommendation that associates with the plurality of object to be predicted
Throw in priority.Owing to, in said process, the history access record of user being split as according to the connected reference behavior of user
Multiple continuous access records, and using continuous access record as the benchmark judged, the access record comprised from continuous access record
In choose positive example sample and negative example sample, such positive example sample and negative example sample and provide model training optimization more accurately
Target, and the regression model obtained based on the training of this target can carry out accurate sequence to object to be predicted so that this row
Sequence can accurately represent user and click on the probability order of object to be predicted, when pushing away of throwing according to this order really directional user
When recommending the priority of content, the probability that user clicks on the higher content recommendation of current priority is the biggest, thus improves
The accuracy that content recommendation is thrown in.Content recommendation owing to throwing in is more accurate, thus can preferably avoid user is acquisition
Its content interested and repeatedly search for and browse etc. operates, and then can save the access again for responding user or search
The network system resources of Suo Suoxu.Simultaneously as the probability that user obtains its content interested from content recommendation increases, from
And reduce the probability of operations such as need to repeatedly searching for and browse, and then improve Consumer's Experience.
Accompanying drawing explanation
Fig. 1 is the flow chart of the put-on method of a kind of internet content in the embodiment of the present invention；
Fig. 2 is the structural representation of the put-on method of a kind of internet content in the embodiment of the present invention.
Detailed description of the invention
As background technology is sayed, the user predicted according to the method for the prior art preferential click order to content recommendation
Not accurate enough, in other words, the content of relatively preferential recommendation can not represent the content that user priority is clicked on exactly.Therefore, use
Family is to obtain its content interested, generally requires operations such as repeatedly searching for and browse, for again browsing of response user
Or search, need to provide more network system resources, cause cost increase.Meanwhile, Consumer's Experience is the most poor.
The technical scheme of the embodiment of the present invention by obtaining active user in the history access record data of current site, and
It is multiple continuous access records by described history access record data cutting, the access comprised from the plurality of continuous access record
Record is chosen positive example sample and negative example sample, each positive example sample and negative example sample is concluded and accesses characteristic, based on
The described access characteristic being marked as positive example or negative example carries out model training, to obtain multiple regression model, then passes through back
Updating of model and object to be predicted is returned to access characteristic and determine the content recommendation that associates with the plurality of object to be predicted
Throw in priority.Owing to, in said process, the history access record of user being split as according to the connected reference behavior of user
Multiple continuous access records, and using continuous access record as the benchmark judged, the access record comprised from continuous access record
In choose positive example sample and negative example sample, such positive example sample and negative example sample and provide model training optimization more accurately
Target, and the regression model obtained based on the training of this target can carry out accurate sequence to object to be predicted so that this row
Sequence can accurately represent user and click on the probability order of object to be predicted, when pushing away of throwing according to this order really directional user
When recommending the priority of content, the probability that user clicks on the higher content recommendation of current priority is the biggest, thus improves
The accuracy that content recommendation is thrown in.Content recommendation owing to throwing in is more accurate, thus can preferably avoid user is acquisition
Its content interested and repeatedly search for and browse etc. operates, and then can save the access again for responding user or search
The network system resources of Suo Suoxu.Simultaneously as the probability that user obtains its content interested from content recommendation increases, from
And reduce the probability of operations such as need to repeatedly searching for and browse, and then improve Consumer's Experience.
Understandable for enabling the above-mentioned purpose of the present invention, feature and beneficial effect to become apparent from, below in conjunction with the accompanying drawings to this
The specific embodiment of invention is described in detail.
Fig. 1 is the flow chart of the put-on method of a kind of internet content in the embodiment of the present invention.Shown in Fig. 1
Step illustrate.
Step S101: obtain active user's history access record data in current site, and described history is accessed note
Record data cutting is multiple continuous access records, and each described continuous access record includes that described active user is to described current net
At least one access record that the connected reference behavior stood is formed, every described access record includes that described active user is described
Access behavioral data to object in current site.
Wherein, current site is the website of internet content to be put, and it can be any suitable website, such as news
Website, video website, shopping website etc..
The history access record data of current site come from the server of described current site, described history access record
Data can be limited with a time range before certain time point, and the access behavioral data of such as above month is as this
History access record data described in embodiment.
In being embodied as, described continuous access record includes that the continuous print of user accesses at least one that behavior is formed
Access record.Wherein, it can be the access behavior during a session (session) that continuous print accesses behavior, such as, from
Family U opens the process of website S to the S that shuts down web sites and is a described user U conversation procedure at this website S.Every described visit
Ask that record includes the described active user access behavioral data to object in described current site.Specifically, every accesses note
Record can include the access time, access object, whether first time by current sessions and access and access type is constituted access
Behavioral data." object " of embodiment of the present invention indication can be news category, video classification, commodity or other are the most right
As.Described access type includes browsing and searching for.When described object is commodity, described access type can also include collection and
Add shopping cart.Illustrate at the part access behavior data instance of website S with user U in table 1 below:
Table 1
In table 1, the access behavioral data of the session that Ts1 to Ts3 defines is the first continuous access record S1, including 3 visits
Ask record.The session access behavioral data that Ts4 and Ts5 defines is the second continuous access record S2, accesses record including 2.Often
Bar access record include the access time, whether be access the first time in current sessions, the commodity that access and access type.Need
Illustrate, access behavioral data and be not limited to above-mentioned enumerating.
Step S102: choose positive example sample and negative example sample from the access record that the plurality of continuous access record comprises
This.
In being embodied as, for the access record in each continuous access record, if described access recording needle pair
Object continuous access record previously is accessed, is then described positive example sample by described access recording mark；For often
Access record in one continuous access record, if quilt in the continuous access record that the object of described access recording needle pair is previously
Accessed, but not accessed in existing continuous accesses record, be then described negative example sample by described access recording mark.
Continue to illustrate the choosing, it is assumed herein that the first continuous access is recorded as of positive example sample and negative example sample with table 1
Selected first continuous access record in history access record data, in other words, history access record data from time
Between Ts1 start.Wherein:
For the first continuous access record S1, any commodity are the most not visited, and therefore first continues
Access in record S1 and do not produce positive example sample and negative example sample.For the second continuous access record S2, due to commodity Pid1
Access S1 in record first to be accessed, and be also accessed in existing continuous accesses record S2, therefore commodity Pid1 pair
This access record answered is marked as positive example sample；Commodity Pid2 is accessed in the first continuous access record S1, but
Being the most accessed in current second continuous access record S2, this therefore corresponding for commodity Pid2 access record is labeled
For negative example sample.
Similarly, for other continuous access records in selected history access record data, it is also possible to Cong Zhongxuan
Go out positive example sample and negative example sample.
It should be noted that when labelling positive example sample and negative example sample, can be not only to be browsed or to be searched for as bar
Part, it is also possible to limit to meet and preset access type as condition.Such as, when described object is commodity, and described access type includes
Browse, search for, when collecting and add shopping cart, existing continuous accessed the access record before record and existing continuous accesses note
In record access type be the object tag of preset kind be described positive example sample, by existing continuous access record before access note
In record, access type is preset kind, but the object tag of access type Non-precondition type is in existing continuous accesses record
Described negative example sample, described preset kind includes search and browses.When choosing positive example sample and negative example sample, need by specifying
The preset kind met, can select the set being considered valuable sample as model training.
Step S103: according to described positive example sample and described negative example sample, visit from described history access record Data induction
Ask characteristic.
In being embodied as, described access characteristic include active user access described current site characteristic,
Object is accessed for characteristic and described active user accesses the characteristic of described object.Wherein, active user accesses institute
The characteristic stating current site can include that active user accesses the statistical data of behavior to current site, such as, can be
Active user accesses the number of times of current site in predetermined period, browses the number of times of different object, the number of times of the different object of search
Deng；Object is accessed for characteristic can include the statistical data of different object, such as, can be that different object is held at each
The number of times browsed in continuous access record, searched number of times etc.；Active user accesses the characteristic of described object
Active user accesses the statistical data of behavior to different objects, such as, can be that active user is in each continuous access record
To the different numbers of visits of object, searching times and last access time etc..
Illustrating as a example by the positive example sample selected from table 1 by step S102 and negative example sample, described object is for be still
Commodity.
Described active user accesses before the characteristic of described current site may include that Ts4, and active user U accesses
The total degree of current site S is 3 times, and the access time last for active user U is Ts3, etc..
Described object is accessed for characteristic and may include that based on commodity Pid1 statistics, before Ts4, commodity Pid1
Accessed totally 2 times (i.e. search for 1 time and browse 1 time)；Add up based on commodity Pid2, before Ts4, commodity Pid2 accessed totally 1
Secondary (i.e. search 1 time).
Described active user U accesses the characteristic of described object and may include that before Ts4, and active user U is to commodity
Pid1 accesses 2 times (i.e. search for 1 time and browse 1 time)；Before Ts4, commodity Pid2 is accessed and (i.e. searches for 11 time by active user U
Secondary)；Before Ts4, active user U is Ts3 to the last access time of commodity Pid1, and active user U is to commodity Pid2
The rear access time is Ts2.
It should be noted that described access characteristic can also be concluded according to other preset standard, at this not
Limit.
Step S104: carry out model training based on described access characteristic, to obtain multiple regression model.
In being embodied as, described regression model can be that gradient promotes decision tree (Gradient Boosting
Decision Tree, GBDT) model.The process of described model training is those skilled in the art's prior aries to understand,
It is not repeated herein.
Step S105: renewals based on multiple objects to be predicted access characteristic and the plurality of regression model, really
The input priority of the fixed content recommendation associated with the plurality of object to be predicted.
Described object to be predicted is the candidate target that may carry out commending contents for active user.For example, it is desired to from 100
User is thrown in by the content recommendation selecting 8 object associations in individual candidate target, and these 100 candidate targets are to be predicted right
As.
In being embodied as, renewal based on object to be predicted can access characteristic and the plurality of regression model and obtain
To the score of described object to be predicted, be ranked up by from high to low further according to described score, and according to described sequence determine with
The input priority of the content recommendation of the plurality of object to be predicted association.Such as, object to be predicted has the first object and
Second object, the access characteristic and the plurality of regression model that update according to the first object and the second object carry out pre-
Surveying, if the score obtaining proportion by subtraction the second object of the first object is high, then in both, the preferential content recommendation throwing in the first object is extremely worked as
Front user.
In being embodied as, can choose and throw in the recommendation associated by object to be predicted that priority orders ranking is the highest
Hold and throw in, or choose and throw in priority orders and be in the content recommendation associated by object to be predicted of top N and throw
Put.Such as, renewals based on 100 objects to be predicted access characteristic and the plurality of regression model, can obtain every
The score of individual object to be predicted, according to score from high to low order, chooses the recommendation of the object to be predicted association of highest scoring
Content is thrown in user, or the content recommendation choosing the object to be predicted association that score is in first 8 is carried out to user
Throw in.Such as, when described object is news category, described content recommendation can be news content；When described object is commodity
Time, described content recommendation can be Internet advertising.
It should be noted that step S101 of the embodiment of the present invention to step S105 is for the active user concrete with
For, same embodiment can be used to determine the content recommendation for this user for accessing other users of current site
Throw in priority.
In embodiments of the invention, accessing the positive example sample chosen record and bearing included in continuous access record
Example sample provides the target that model training optimizes, and can be to object to be predicted based on this target regression model of obtaining of training
Carry out accurate sequence so that this sequence can accurately represent user and click on the probability order of object to be predicted, and then
Promote the accuracy estimating the probability that user clicks on content recommendation, when according to this order content recommendation that really directional user throws in
Priority time, it is the biggest that user clicks on the probability of the higher content recommendation of current priority, thus preferably avoids using
Family be obtain its content interested repeatedly search for and browse wait operation, and then can save into response user visit again
Ask or search for required network system resources.Simultaneously as user obtains the possibility of its content interested from content recommendation
Property increase, thus reduce need to repeatedly search for and browse and wait the probability operated, and then improve Consumer's Experience.
Fig. 2 is the structural representation of the delivery device of a kind of internet content in the embodiment of the present invention.As shown in Figure 2
The delivery device of internet content may include that acquiring unit 201, chooses unit 202, conclusion unit 203, model training unit
204 and determine unit 205.
Described acquiring unit 201, is suitable to the history access record data obtaining active user in current site, and by described
History access record data cutting is multiple continuous access records, and each described continuous access record includes described active user couple
At least one access record that the connected reference behavior of described current site is formed, every described access record include described currently
User's access behavioral data to object in described current site.
Described choose unit 202, be suitable to choose positive example sample from the access record that the plurality of continuous access record comprises
Originally with negative example sample.
In being embodied as, described in choose unit 202 and may include that
First labelling subelement, is suitable to for the access record in each continuous access record, if described access record
For object continuous access record previously in be accessed, then be described positive example sample by described access recording mark；
Second labelling subelement, is suitable to for the access record in each continuous access record, if described access record
For object continuous access record previously in be accessed, but the most accessed, then in existing continuous accesses record
It is described negative example sample by described access recording mark.
Described conclusion unit 203, is suitable to according to described positive example sample and described negative example sample, from described history access record
Data induction accesses characteristic.
Described model training unit 204, is suitable to carry out model training based on described access characteristic, to obtain multiple times
Return model.
In being embodied as, described regression model can be GBDT model.
Described determine unit 205, be suitable to renewal based on multiple objects to be predicted and access characteristic and the plurality of
Regression model, determines the input priority of the content recommendation associated with the plurality of object to be predicted.
In being embodied as, described determine that unit 205 may include that
Score obtains subelement, is suitable to renewal based on object to be predicted and accesses characteristic and the plurality of regression model
Obtain the score of described object to be predicted；
Release sequence determines subelement, is suitable to be ranked up by from high to low according to described score, with according to described sequence
Determine the input priority of the content recommendation associated with the plurality of object to be predicted.
Structure explanation about the delivery device of described internet content can be to should refer to described in Fig. 1 with beneficial effect
The enforcement explanation of the put-on method of internet content and beneficial effect, do not repeat them here.
In being embodied as, when the put-on method of described internet content can apply to Internet advertising field, institute
The delivery device stating internet content can apply to DSP server.
One of ordinary skill in the art will appreciate that all or part of step in the various methods of above-described embodiment is can
Completing instructing relevant hardware by program, this program can be stored in computer-readable recording medium, and storage is situated between
Matter may include that ROM, RAM, disk or CD etc..
Having been described in detail the method and system of the embodiment of the present invention above, the present invention is not limited to this.Any
Skilled person, without departing from the spirit and scope of the present invention, all can make various changes or modifications, therefore the guarantor of the present invention
The scope of protecting should be as the criterion with claim limited range.
Claims (10)
1. the put-on method of an internet content, it is characterised in that including:
Obtain active user's history access record data in current site, and be many by described history access record data cutting
Individual continuous access record, each described continuous access record includes the described active user connected reference row to described current site
For at least one access record formed, every described access record include described active user in described current site to right
The access behavioral data of elephant；
Positive example sample and negative example sample is chosen from the access record that the plurality of continuous access record comprises；
According to described positive example sample and described negative example sample, access characteristic from described history access record Data induction；
Model training is carried out, to obtain multiple regression model based on described access characteristic；
Renewals based on multiple objects to be predicted access characteristic and the plurality of regression model, determine and treat with the plurality of
The input priority of the content recommendation of prediction object association.
The put-on method of internet content the most according to claim 1, it is characterised in that remember from the plurality of continuous access
Positive example sample chosen in the access record that record comprises and negative example sample include:
For the access record in each continuous access record, if the lasting visit that the object of described access recording needle pair is previously
Ask in record and be accessed, be then described positive example sample by described access recording mark；
For the access record in each continuous access record, if the lasting visit that the object of described access recording needle pair is previously
Ask in record and be accessed, but not accessed in existing continuous accesses record, be then described by described access recording mark
Negative example sample.
The put-on method of internet content the most according to claim 1, it is characterised in that described regression model is GBDT tree
Model.
The put-on method of internet content the most according to claim 1, it is characterised in that based on multiple objects to be predicted
Update and access characteristic and the plurality of regression model, determine the content recommendation that associates with the plurality of object to be predicted
Throw in priority, including:
Renewal based on object to be predicted accesses characteristic and the plurality of regression model obtains obtaining of described object to be predicted
Point；
It is ranked up by from high to low according to described score, and determines according to described sequence and associate with the plurality of object to be predicted
The input priority of content recommendation.
The put-on method of internet content the most according to claim 1, it is characterised in that described access characteristic includes
Active user accesses the characteristic of described current site, object is accessed for characteristic and described active user accesses described
The characteristic of object.
6. the delivery device of an internet content, it is characterised in that including:
Acquiring unit, is suitable to the history access record data obtaining active user in current site, and described history is accessed note
Record data cutting is multiple continuous access records, and each described continuous access record includes that described active user is to described current net
At least one access record that the connected reference behavior stood is formed,
Every the described record that accesses includes the described active user access behavioral data to object in described current site；
Choose unit, be suitable to from the access record that the plurality of continuous access record comprises, choose positive example sample and negative example sample
This；
Conclude unit, be suitable to, according to described positive example sample and described negative example sample, visit from described history access record Data induction
Ask characteristic；
Model training unit, is suitable to carry out model training based on described access characteristic, to obtain multiple regression model；
Determine unit, be suitable to renewal based on multiple objects to be predicted and access characteristic and the plurality of regression model, really
The input priority of the fixed content recommendation associated with the plurality of object to be predicted.
The delivery device of internet content the most according to claim 6, it is characterised in that described in choose unit and include:
First labelling subelement, is suitable to for the access record in each continuous access record, if described access recording needle pair
Object continuous access record previously in be accessed, then be described positive example sample by described access recording mark；
Second labelling subelement, is suitable to for the access record in each continuous access record, if described access recording needle pair
Object continuous access record previously in be accessed, but the most accessed, then by institute in existing continuous accesses record
Stating access recording mark is described negative example sample.
The delivery device of internet content the most according to claim 6, it is characterised in that described regression model is GBDT mould
Type.
The delivery device of internet content the most according to claim 6, it is characterised in that described determine that unit includes:
Score obtains subelement, is suitable to renewal based on object to be predicted access characteristic and the plurality of regression model obtains
The score of described object to be predicted；
Release sequence determines subelement, is suitable to be ranked up by from high to low according to described score, and determines according to described sequence
Input priority with the content recommendation that the plurality of object to be predicted associates.
The delivery device of internet content the most according to claim 6, it is characterised in that described access characteristic bag
Include that active user accesses the characteristic of described current site, object is accessed for characteristic and described active user accesses institute
State the characteristic of object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610424892.7A CN106126544B (en) | 2016-06-15 | 2016-06-15 | Internet content delivery method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610424892.7A CN106126544B (en) | 2016-06-15 | 2016-06-15 | Internet content delivery method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106126544A true CN106126544A (en) | 2016-11-16 |
| CN106126544B CN106126544B (en) | 2020-01-24 |
Family
ID=57469461
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610424892.7A Active CN106126544B (en) | 2016-06-15 | 2016-06-15 | Internet content delivery method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106126544B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107153684A (en) * | 2017-04-24 | 2017-09-12 | 北京小米移动软件有限公司 | Display methods, device and the equipment of PUSH message |
| CN107633326A (en) * | 2017-09-14 | 2018-01-26 | 北京拉勾科技有限公司 | A kind of user delivers the construction method and computing device of wish model |
| WO2018126953A1 (en) * | 2017-01-06 | 2018-07-12 | 腾讯科技（深圳）有限公司 | Seed population expanding method, device, information releasing system and storing medium |
| CN108304853A (en) * | 2017-10-10 | 2018-07-20 | 腾讯科技（深圳）有限公司 | Acquisition methods, device, storage medium and the electronic device for the degree of correlation of playing |
| CN110674434A (en) * | 2019-09-26 | 2020-01-10 | 秒针信息技术有限公司 | Method and device for releasing browsing resources |
| CN114417817A (en) * | 2021-12-30 | 2022-04-29 | 中国电信股份有限公司 | Session information cutting method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100329545A1 (en) * | 2009-06-30 | 2010-12-30 | Xerox Corporation | Method and system for training classification and extraction engine in an imaging solution |
| CN103870972A (en) * | 2012-12-07 | 2014-06-18 | 盛乐信息技术（上海）有限公司 | Data recommendation method and data recommendation system |
| CN104166668A (en) * | 2014-06-09 | 2014-11-26 | 南京邮电大学 | News recommendation system and method based on FOLFM model |
| CN105631538A (en) * | 2015-12-23 | 2016-06-01 | 北京奇虎科技有限公司 | User activity prediction method and device, and application method and system thereof |
-
2016
- 2016-06-15 CN CN201610424892.7A patent/CN106126544B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100329545A1 (en) * | 2009-06-30 | 2010-12-30 | Xerox Corporation | Method and system for training classification and extraction engine in an imaging solution |
| CN103870972A (en) * | 2012-12-07 | 2014-06-18 | 盛乐信息技术（上海）有限公司 | Data recommendation method and data recommendation system |
| CN104166668A (en) * | 2014-06-09 | 2014-11-26 | 南京邮电大学 | News recommendation system and method based on FOLFM model |
| CN105631538A (en) * | 2015-12-23 | 2016-06-01 | 北京奇虎科技有限公司 | User activity prediction method and device, and application method and system thereof |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018126953A1 (en) * | 2017-01-06 | 2018-07-12 | 腾讯科技（深圳）有限公司 | Seed population expanding method, device, information releasing system and storing medium |
| US11301884B2 (en) | 2017-01-06 | 2022-04-12 | Tencent Technology (Shenzhen) Company Limited | Seed population diffusion method, device, information delivery system and storage medium |
| CN107153684A (en) * | 2017-04-24 | 2017-09-12 | 北京小米移动软件有限公司 | Display methods, device and the equipment of PUSH message |
| CN107153684B (en) * | 2017-04-24 | 2020-02-07 | 北京小米移动软件有限公司 | Display method, device and equipment of push message |
| CN107633326A (en) * | 2017-09-14 | 2018-01-26 | 北京拉勾科技有限公司 | A kind of user delivers the construction method and computing device of wish model |
| CN108304853A (en) * | 2017-10-10 | 2018-07-20 | 腾讯科技（深圳）有限公司 | Acquisition methods, device, storage medium and the electronic device for the degree of correlation of playing |
| CN110674434A (en) * | 2019-09-26 | 2020-01-10 | 秒针信息技术有限公司 | Method and device for releasing browsing resources |
| CN114417817A (en) * | 2021-12-30 | 2022-04-29 | 中国电信股份有限公司 | Session information cutting method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106126544B (en) | 2020-01-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106126544A (en) | The put-on method of a kind of internet content and device | |
| Terrance et al. | Importance of Search Engine Marketing in the Digital World. | |
| RU2696230C2 (en) | Search based on combination of user relations data | |
| US9268826B2 (en) | System and method for crowdsourced template based search | |
| US20190018900A1 (en) | Method and Apparatus for Displaying Search Results | |
| CN105765573B (en) | Improvements in website traffic optimization | |
| US20160179953A1 (en) | Time-Box Constrained Searching in a Distributed Search System | |
| US8682882B2 (en) | System and method for automatically identifying classified websites | |
| US20170154116A1 (en) | Method and system for recommending contents based on social network | |
| EP2159716A1 (en) | System and method for interfacing a web browser widget with social indexing | |
| US20100169756A1 (en) | Automated bookmarking | |
| US8380707B1 (en) | Session-based dynamic search snippets | |
| US20130212089A1 (en) | Search Result Categorization | |
| CN102841904B (en) | A kind of searching method and equipment | |
| US10853424B1 (en) | Content delivery using persona segments for multiple users | |
| JP2015537319A (en) | Method and system for pushing mobile applications | |
| CN102708174A (en) | Method and device for displaying rich media information in browser | |
| CN105224554A (en) | Search word is recommended to carry out method, system, server and the intelligent terminal searched for | |
| JP6966158B2 (en) | Methods, devices and programs for processing search data | |
| CN101751422A (en) | Method, mobile terminal and server for carrying out intelligent search at mobile terminal | |
| US9405803B2 (en) | Ranking signals in mixed corpora environments | |
| CN106033415A (en) | A text content recommendation method and device | |
| CN102819384B (en) | A kind of input field place carries out the method and apparatus of pointing out display | |
| CN107066509A (en) | Information providing method, processing server and merging server | |
| CN105630937A (en) | Method and device for searching answers to exam questions |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |