CN110222975A - A kind of loss customer analysis method, apparatus, electronic equipment and storage medium - Google Patents
A kind of loss customer analysis method, apparatus, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN110222975A CN110222975A CN201910473357.4A CN201910473357A CN110222975A CN 110222975 A CN110222975 A CN 110222975A CN 201910473357 A CN201910473357 A CN 201910473357A CN 110222975 A CN110222975 A CN 110222975A
- Authority
- CN
- China
- Prior art keywords
- user
- target
- characteristic
- target user
- lost
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- 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
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Accounting & Taxation (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention provides a kind of loss customer analysis method, apparatus, electronic equipment and storage mediums, and wherein method comprises determining that target user, and extracts the characteristic of target user;The characteristic of target user is input to the decision Tree algorithms model constructed in advance, judges whether target user is to be lost user;When target user is to be lost user, based on the characteristic of target user, target user is evaluated by default scoring tactics, obtain evaluation result corresponding to each default scoring tactics, and the evaluation result according to corresponding to obtained default scoring tactics, determine that the target of target user is lost type;Determine that target is lost that type is corresponding keeps strategy, so that keeping target user by keeping strategy.The present invention realizes analysis user behavior, it is determined whether to be lost user, analysis is lost the loss type of user, targetedly formulates and keeps strategy, and success rate is kept in promotion, cuts operating costs.
Description
Technical field
The present invention relates to data analysis technique fields, set more particularly to a kind of loss customer analysis method, apparatus, electronics
Standby and storage medium.
Background technique
With the rapid development of video website business, the user of video website flows in continuous growth or within certain time
It loses.Customer churn can cause adverse effect to enterprise, such as reduce enterprise's income, influence enterprise Institutions;Reduce enterprise income rate;
It improves enterprise marketing and user recalls cost.In face of the change and fierce market competition of user watched habit, customer churn is
Develop most severe one of pressure as video website.
Newly-increased the same with user, customer churn may all occur all the time, and excavate unregistered user's needs
Cost be to keep six times of registered user here, therefore how to keep registered user here and ask as technology urgently to be resolved
Topic.
Summary of the invention
Be designed to provide a kind of loss customer analysis method, apparatus, electronic equipment and the storage of the embodiment of the present invention are situated between
Matter, to realize analysis user behavior, it is determined whether to be lost user, and analysis is lost the loss type of user, targetedly
Strategy is kept in formulation, and promotion keeps success rate, cuts operating costs.Specific technical solution is as follows:
In a first aspect, the embodiment of the invention discloses a kind of loss customer analysis methods, which comprises
It determines target user, and extracts the characteristic of the target user;
The characteristic of the target user is input to preset decision Tree algorithms model, judges that the target user is
No is to be lost user;
When the target user is to be lost user, based on the characteristic of the target user, commented by default
Point strategy evaluates the target user, obtains evaluation result corresponding to each default scoring tactics, and according to
Evaluation result corresponding to the obtained default scoring tactics determines that the target of the target user is lost type;
Determine that the target is lost that type is corresponding to keep strategy so that by it is described keep strategy and keep the target use
Family.
Optionally, in the determining target user, and after extracting the characteristic of the target user, the method is also
Include:
Data prediction is carried out to the characteristic of the target user, obtains feature needed for analyzing the target user
Data, as target signature data;
According to preset rules, based on the target signature data of the target user, judge that the target user is new
User or old user;
The characteristic by the target user is input to preset decision Tree algorithms model, judges that the target is used
Whether family is to be lost user, comprising:
When the target user is new user, the characteristic of the target user is input to preset new user and is determined
Plan tree algorithm model judges whether the target user is to be lost user;
When the target user is old user, the characteristic of the target user is input to preset old user and is determined
Plan tree algorithm model judges whether the target user is to be lost user.
Optionally, the step of constructing the preset decision Tree algorithms model, comprising:
The history feature data for choosing multiple users in preset time period, as each sample characteristics data;
According to preset rules, each sample characteristics data are divided into new user's history characteristic or old user respectively
History feature data;
To each new user's history characteristic and each old user's history feature data, data prediction is carried out,
Characteristic needed for analyzing each new user is obtained, as needed for new ownership goal history feature data and each old user of analysis
Characteristic, as old user's target histories characteristic;
By each new ownership goal history feature data, using decision Tree algorithms, it is default accurate that training obtains meeting
New user's decision Tree algorithms model of rate;
By each old user's target histories characteristic, using decision Tree algorithms, training obtains meeting described default
Old user's decision Tree algorithms model of accuracy rate.
Optionally, the scoring tactics are that whether active in the recent period, frequency of use height and video playing duration be long with user
What the short analysis foundation for being characterized data was established.
Second aspect, the embodiment of the invention discloses a kind of loss customer analysis device, described device includes:
Characteristic extraction module for determining target user, and extracts the characteristic of the target user;
It is lost user's judgment module, for the characteristic of the target user to be input to preset decision Tree algorithms mould
Type judges whether the target user is to be lost user;
Target is lost determination type module, is used for when the target user is to be lost user, with the target user's
Based on characteristic, the target user is evaluated by default scoring tactics, obtains each default scoring plan
Slightly corresponding evaluation result, and the evaluation result according to corresponding to the obtained default scoring tactics, determine the mesh
The target for marking user is lost type;
Tactful determining module is kept, for determining that the target is lost that type is corresponding to keep strategy, so that by described
It keeps strategy and keeps the target user.
Optionally, described device further include:
Target signature data determining module carries out data prediction for the characteristic to the target user, obtains
Characteristic needed for analyzing the target user, as target signature data;
Target user's division module is used for according to preset rules, based on the target signature data of the target user,
Judge the target user for new user or old user;
Loss user's judgment module, comprising:
First-class appraxia family judging submodule is used for when the target user is new user, by the target user's
Characteristic is input to preset new user's decision Tree algorithms model, judges whether the target user is to be lost user;
Second appraxia family judging submodule is used for when the target user is old user, by the target user's
Characteristic is input to preset old user's decision Tree algorithms model, judges whether the target user is to be lost user.
Optionally, described device further include:
Sample characteristics data determining module, for choosing the history feature data of multiple users in preset time period, as
Each sample characteristics data;
Sample characteristics data division module, for being respectively divided into each sample characteristics data according to preset rules
New user's history characteristic or old user's history feature data;
Target histories characteristic determining module, for each new user's history characteristic and each old user
History feature data carry out data prediction, characteristic needed for analyzing each new user, as new ownership goal history feature
Characteristic needed for data and each old user of analysis, as old user's target histories characteristic;
New user's decision Tree algorithms model determining module, for passing through each new ownership goal history feature data, benefit
With decision Tree algorithms, training obtains the new user's decision Tree algorithms model for meeting default accuracy rate;
Old user's decision Tree algorithms model determining module, for passing through each old user's target histories characteristic, benefit
With decision Tree algorithms, training obtains the old user's decision Tree algorithms model for meeting the default accuracy rate.
Optionally, the scoring tactics are that whether active in the recent period, frequency of use height and video playing duration be long with user
What the short analysis foundation for being characterized data was established.
The third aspect, the embodiment of the invention discloses a kind of electronic equipment, including processor, communication interface, memory and
Communication bus, wherein the processor, the communication interface, the memory are completed each other by the communication bus
Communication;
The memory, for storing computer program;
The processor when for executing the program stored on the memory, realizes above-mentioned loss customer analysis side
Any method and step in method.
At the another aspect that the present invention is implemented, a kind of computer readable storage medium is additionally provided, it is described computer-readable
It is stored with instruction in storage medium, when run on a computer, realizes any described in above-mentioned loss customer analysis method
Method and step.
At the another aspect that the present invention is implemented, the embodiment of the invention also provides a kind of, and the computer program comprising instruction is produced
Product realize any method and step in above-mentioned loss customer analysis method when run on a computer.
A kind of loss customer analysis method, apparatus, electronic equipment and storage medium provided in an embodiment of the present invention, Ke Yifen
User behavior is analysed, specifically, determining target user, and extracts the characteristic of target user;It is excavated in the embodiment of the present invention
One more complete characteristic set is used to describe the behavior of user.And then the characteristic of target user is input to preparatory structure
The decision Tree algorithms model built judges whether target user is to be lost user;It, can be accurate using the disaggregated model of tree class
Characteristic is analyzed, judges whether target user is to be lost user.When prejudging out the target user to be lost user, further
User group belonging to the loss user of anticipation will be determined by scoring tactics, it is finally targetedly real for the user group
Row keeps strategy, promotes the success rate of keeping of registered users, can also cut operating costs.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described.
Fig. 1 is a kind of loss customer analysis method flow diagram of the embodiment of the present invention;
Fig. 2 be the embodiment of the present invention a kind of loss customer analysis method in judge target user whether be lost user
Method flow diagram;
Fig. 3 is decision Tree algorithms method for establishing model process in a kind of loss customer analysis method of the embodiment of the present invention
Figure;
Fig. 4 is each sample characteristics data observation period schematic diagram in a kind of loss customer analysis method of the embodiment of the present invention;
Fig. 5 is the logic relation picture of new user and old user in a kind of loss customer analysis method of the embodiment of the present invention;
Fig. 6 is that new user compares with the loss accounting of old user in a kind of loss customer analysis method of the embodiment of the present invention
Figure;
Fig. 7 is a kind of loss customer analysis method logical framework figure of the embodiment of the present invention;
Fig. 8 is a kind of loss customer analysis apparatus structure schematic diagram of the embodiment of the present invention;
Fig. 9 is a kind of electronic equipment structural schematic diagram of the embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention is described.
By the characteristic of acquisition user in the embodiment of the present invention, and the spy of the disaggregated model analysis user using tree class
Data are levied, judge whether user is to be lost user, and analysis is lost the loss type of user, the later period can be helped targetedly to make
Surely strategy is kept, promotion keeps success rate, cuts operating costs.Concrete scheme is as follows:
In a first aspect, the embodiment of the invention discloses a kind of loss customer analysis methods, as shown in Figure 1.Fig. 1 is the present invention
A kind of loss customer analysis method flow diagram of embodiment, method include:
S101 determines target user, and extracts the characteristic of target user.
In the embodiment of the present invention, target user is any user in set period, and video website can be to set period
Interior each user is performed both by the loss customer analysis method of the embodiment of the present application, so that it is determined that out in set period each user whether be
It is lost user and is lost type.It specifically can refer to fix the date as observational day, extract the history number before the observational day in set period
Characteristic in.By this feature data, judge using the observational day as the target user in the following set period of starting point
It whether is to be lost user.Being lost user is the use using the observational day not enliven behavior in the following set period of starting point
Family.
For example, set period is 30 days;Scheduled date is 2019.2.1, then can extract 2019.1.2- in this step
2019.1.31 the characteristic of target user, by the characteristic judgement future of the set period with 2019.2.1 starting point
30 days futures the target user whether be lost user.
The essential information of user is mainly acquired in the embodiment of the present invention, enlivens behavior, broadcasting behavior as characteristic.Example
Such as, user's gender, at the age, last be actively spaced number of days, forecast date twice and finally enliven date intervals number of days, whether step on
Record, whether member, member's rank, the order amount of money, the History Order amount of money, positive playback volume, playing duration, each channel in 6 months
Playback volume, plays video total amount, the number of starts at the number of sessions without the behavior of broadcasting, and positive playback volume, enlivens day at playing duration
Number of days etc. is used as characteristic.
The characteristic of target user is input to the decision Tree algorithms model constructed in advance, judges target user by S102
It whether is to be lost user.
Decision Tree algorithms are a kind of methods for approaching discrete function value.It is a kind of typical classification method, first logarithm
According to being handled, readable rule and decision tree are generated using inductive algorithm, then new data is analyzed using decision.This
Decision tree is the process classified by series of rules to data in matter.
In the embodiment of the present invention, decision Tree algorithms model is constructed in advance by decision Tree algorithms, pass through decision Tree algorithms mould
The characteristic of type analysis target user judges whether the target user is to be lost user.Specific building decision Tree algorithms model
Method following embodiment in be described in detail.
For example, promoting decision-tree model XGBoost (eXtreme using the gradient constructed in advance in the embodiment of the present invention
Gradient Boosting), the characteristic of target user is analyzed, judges whether the target user is to be lost user.
S103, when target user is determined as being lost user, based on the characteristic of the target user, by default
Scoring tactics evaluate target user, obtain evaluation result corresponding to each default scoring tactics, and according to acquired
Evaluation result, determine the target user target be lost type.
In one embodiment, when target user is to be lost user, based on the characteristic of target user, respectively
It is target user's marking by the default grade form of the every kind of type that scores, the scoring type of the highest default grade form of score is true
The target for being set to target user is lost type;Any default grade form is the scoring item determined according to specified scoring type
Table.
After identification is lost user, implementing to keep to it is the most important thing in operation, the considerations of in view of operation cost,
It prefers to be to keep valuable user here.No matter if how product content changes user will not reattempt use, such as
The reasons such as binding, prepackage enter, and only contribute to new increment, do not have more intersections with product, this kind of user keep it is relatively difficult, that
More resources are launched on keeping user also not act on;In addition, how to accomplish there is friendly use when implementing and keeping
Family experience needs to segment the object kept, and push user is interested, positions the content being consistent with user psychology, so
Can maximize avoids user from disliking, and could constantly excite the interest of user, promote the success rate kept.
Therefore, convection current appraxia family is needed further to excavate their stream after identification is lost user in the embodiment of the present invention
Lose reason.Passed through in embodiments of the present invention by the different types of default grade form pre-established to be lost user's marking
Type analysis determines Drain Causes, specifies convenient for the later period and corresponding keeps strategy.
It in the embodiment of the present invention, can classify according to the attribute of user characteristic data, and then obtain each default scoring class
Type constructs the default grade form of the default scoring type for every kind of scoring type respectively.
Optionally, method further include:
The analysis base of data characterized by, frequency of use height whether active in the recent period by user and video playing duration length
Plinth pre-establishes each default grade form of default scoring type.
Concretely:
Step 1, whether active in the recent period, frequency of use height and video playing duration length by user, as to user's
The scoring reference index that characteristic is analyzed;
The scoring item of every kind of scoring reference index is arranged for every kind of scoring reference index in step 2;
Multiple expression ranges of characteristic and every kind is arranged under the scoring item for each scoring item in step 3
The corresponding score value of expression range;
Step 4, by an expression range of every kind of scoring item and default the commenting of the corresponding score value formation of the expression range
Divide table, be determined as a type of default grade form, makes the default grade form for establishing every kind of scoring type.
For example, according to the popularization demand of operational diversification, user that prediction is lost according to whether it is active in the recent period, use frequency
Three aspects of rate height and playing duration length form 4 as the scoring reference index that the characteristic to user is analyzed
The default grade form of kind scoring type, the performance of the default grade form of every kind of scoring type is respectively:
1 class of high value presets user's performance of grade form: it is non-it is active in the recent period, frequency of use is low, playing duration;
2 class of high value presets user's performance of grade form: enlivening in the recent period, frequency of use is low, playing duration is long;
1 class of low value presets user's performance of grade form: it is non-it is active in the recent period, frequency of use is low, playing duration is short;
2 class of low value presets user's performance of grade form: enlivening in the recent period, frequency of use is low, playing duration is short;
The scoring item that the default grade form of every kind of scoring type includes can are as follows: observational day distance finally enlivens day number of days;
Finally actively it is separated by number of days twice;The number of starts of day is finally enlivened before forecast date;Broadcasting for day is finally enlivened before forecast date
Put duration;The broadcasting number of videos of day is finally enlivened before forecast date;The positive rate of day is finally enlivened before forecast date;Predict day
The broadcasting channel number of day was finally enlivened before phase;The number of starts before forecast date in 7 days;When broadcasting before forecast date in 7 days
It is long.
Such as 1 class user's expert analysis mode table of high value shown in table 1.
Table 1
In the embodiment of the present invention, the default grade form and above-mentioned 1 class user's expert analysis mode table of high value of other scoring types
In each scoring item it is identical, it is different in performance and score value, specifically repeat no more.
It can be respectively target user's marking by the default grade form of the above-mentioned 4 kinds types that score, score is highest default
The target that the scoring type of grade form is determined as target user is lost type.
For example, target user A, referring to 1 class user's expert analysis mode table score 13 of high value, 2 class of high value is preset grade form and is obtained
Divide 32,1 class of low value presets grade form score 26, and 2 class of low value presets grade form score 19, then the target of the target user
Being lost type is 2 class of high value.
S104 determines that the target is lost that type is corresponding to keep strategy so that by it is described keep strategy keep it is described
Target user.
In the embodiment of the present invention, can in advance according to the setting of scoring type with score that type is corresponding keeps strategy, above-mentioned
It has been determined that the target of target user is lost type, can have been determined in strategy and target loss in a plurality of types of keep pre-established
Type is corresponding to keep strategy, keeps strategy by this and keeps as far as possible to the target user.
For example, the user that above-mentioned 1 class of high value presets grade form show as it is non-it is active in the recent period, frequency of use is low, when playing
Long, then the strategy of keeping of the type user can are as follows: the playing duration of single video is obtained, when being greater than default play for user's push
The video of long same video type;2 class of high value presets user's performance of grade form: active in the recent period, frequency of use is low, plays
Duration is long, then the strategy of keeping of the type user can are as follows: obtains video and video that the nearest preset time period of user was watched
Playing duration, count all types of video playing total durations according to the video type of viewing, pushed for user and play total duration most
The video of the same type of big type.It specifically keeps strategy and system is showed according to the user of every kind of default grade form by implementation personnel
Fixed, the embodiment of the present invention is without limitation.
In addition, can also be lost type according to the target analyzed, formulates keep strategy in real time, kept as far as possible by this
The target user is kept.
In a kind of loss customer analysis method provided in an embodiment of the present invention, specifically, determining target user, and extract
The characteristic of target user;A more complete characteristic set has been excavated in the embodiment of the present invention for describing the row of user
For.And then the characteristic of target user is input to the decision Tree algorithms model constructed in advance, judge target user whether be
It is lost user;Using the disaggregated model of tree class, characteristic can be accurately analyzed, judges whether target user is to be lost to use
Family.When prejudging out the target user to be lost user, will further be determined by scoring tactics belonging to the loss user of anticipation
User group, finally targetedly carry out for the user group and keep strategy, promote registered users keeps success
Rate can also cut operating costs.
Optionally, target user is being determined in a kind of loss customer analysis device of the embodiment of the present invention, and extract target
After the characteristic of user, can have it is shown in Fig. 2 judge target user whether be lost user method flow diagram, method packet
It includes:
S201 carries out data prediction to the characteristic of target user, obtains the target signature data of target user.
Wherein, process of data preprocessing is concretely: carrying out Missing Data Filling, discrete to the characteristic of target user
Change, arithmetic, the treatment process for removing exceptional value.
S202 based on the target signature data of target user, judges target user for new user according to preset rules
Or old user;
The characteristic of target user is input to the decision Tree algorithms model constructed in advance in above-mentioned S102, judges target
Whether user is to be lost user, comprising:
The characteristic of target user is input to predetermined new user when target user is new user by S203
Decision Tree algorithms model judges whether target user is to be lost user;
The characteristic of target user is input to predetermined old user when target user is old user by S204
Decision Tree algorithms model judges whether target user is to be lost user.
Through the embodiment of the present invention, target user has effectively been judged as new user or old user, and then by using with target
Type corresponding decision Tree algorithms model in family judges the loss type of the target user, so that more accurate has judged target use
Whether family is to be lost user.
Optionally, in a kind of embodiment of loss customer analysis method of the invention, decision tree shown in Fig. 3 can calculates
Method method for establishing model flow chart.Decision Tree algorithms model includes new user's decision Tree algorithms model and old user's decision Tree algorithms
Model, predefine decision Tree algorithms model the step of include:
S301 chooses the history feature data of multiple users in preset time period, as each sample characteristics data.
Different types of user will appear before loss and stablize the behavioural characteristic that enliven period different, and these rows
It is characterized window (between winter and summer vacations, popular drama online period and ordinary times) in different times and has different body again
It is existing.So the history feature data of different observation phase all users can be acquired in this step, as each sample characteristics data.Example
Such as, the history feature data of the different observational day previous moons and observational day first trimester are acquired as each sample characteristics data.
For example, choosing on June 1st, 2018 and on July 15th, 2018 in the embodiment of the present invention as forecast date, setting is pre-
If the period is 30, then can be by preset time period May 31-2018 years on the 2nd May in 2018 and-2018 years on the 16th June in 2018
July 14 user history feature data, as each sample characteristics data.Total sample size about 800,000,000 in two periods, it is random to select
Take two periods each 1,200,000 as each sample characteristics data of decision Tree algorithms model, wherein randomly selecting wherein 4/5 conduct
Training dataset, remaining 1/5 is used as test data set.Such as a kind of loss user of the embodiment of the present invention shown in Fig. 4 point
Each sample characteristics data observation period schematic diagram in analysis method, can be special by the history of preset time period user before forecast date in Fig. 4
Data are levied as target user's behavior observation phase, using the history feature data of preset time period user after forecast date as target
The customer churn observation period.
In the embodiment of the present invention from the essential information of user, enliven behavior, broadcasting behavior, collection of drama scale and working days several sides
Face constructs decision Tree algorithms model.The characteristic then extracted in this step can are as follows: user basic information index of correlation: newly-increased day
Number, platform, gender, age;It actively shows index of correlation: being finally actively spaced number of days, forecast date twice and finally enliven day
Period counts every other day;Day behavior index of correlation is finally enlivened before forecast date: whether log in, whether member, member's rank, 6
Month in the order amount of money, the History Order amount of money, all kinds of membership types sum, positive playback volume, playing duration, each channel playback volume,
Positive rate, plays video total amount, the number of starts at the number of sessions without the behavior of broadcasting;It is played in 7 days 3 days before forecast date related
Index: positive playback volume, playing duration, positive rate, average daily positive playback volume;90 days each periods are active before forecast date
Starting distribution index of correlation: number of days, the number of starts are enlivened;90 days each periods actively started distribution correlation and refer to before forecast date
Mark: be lost the observation period whether winter and summer vacation, be lost the observation period in whether have quick-fried play etc..
Each sample characteristics data are divided into new user's history characteristic or old use respectively according to preset rules by S302
Family history feature data.
The history feature data stability of old user and new user are different, and it is also different to be lost trend.Therefore, the present invention is real
It applies and has distinguished new user's history characteristic and old user's history feature data in example, new user's decision is respectively set convenient for the later period
Tree algorithm model and old user's decision Tree algorithms model are targetedly predicted to be lost to be lost in user and old user in new user to use
Family.Fig. 5 is the logic relation picture of new user and old user in a kind of loss customer analysis method of the embodiment of the present invention.
The preset rules of the embodiment of the present invention be Add User data time and pair of forecast date and preset time period
It should be related to.Specifically, detecting in sample data which data that Add User in this step is before forecast date in preset time period
Data, these data are determined as new user's history characteristic, the corresponding user of new user's history characteristic is determined
For new user.By the pervious data that Add User of preset time period before forecast date, it is determined as old user's history feature data, it will
The correspondence user of old user's history feature data is determined as old user.For example, by the data that Add User in front of forecast date 30 days
It is determined as new user's history characteristic, the correspondence user of new user's history characteristic is determined as new user;It will predict day
30 days or more the data that Add User are determined as old user's history feature data before phase, by new user's history characteristic to application
Family is determined as old user.
Fig. 6 is that new user compares with the loss accounting of old user in a kind of loss customer analysis method of the embodiment of the present invention
Figure, the loss accounting of new user is about 2 times of old user as can be known from Fig. 6, illustrates that new user is more easy to run off.The embodiment of the present invention
It is modeled respectively for old and new users Liang Ge group, in conjunction with business scenario and each group's ratio demand, draws a circle to approve old and new users group respectively
The user that middle prediction can be lost.
S303 carries out data prediction, obtains to each new user's history characteristic and each old user's history feature data
Each new ownership goal history feature data and each old user's target histories characteristic.
Wherein, process of data preprocessing is concretely: carrying out Missing Data Filling, discrete to the characteristic of target user
Change, arithmetic, the treatment process for removing exceptional value.
S304, by each new ownership goal history feature data, using decision Tree algorithms, it is default accurate that training obtains meeting
New user's decision Tree algorithms model of rate.
In the embodiment of the present invention, using GBDT (Gradient Boosting Decision Tree, grad enhancement decision
Tree algorithm) two classification are carried out, in specific implementation, the XGBoost method in python/R has been used, previous step is pretreated
Input of each new ownership goal history feature data as model, being lost user's mark is 1, and not being lost user's mark is 0, as
Target class label.Class label vector of the target class label as model, by original data packet, a part is remained as training set
Under part as test set, classifier is trained with training set, recycles test set to test the obtained model of training.
Different parts is repeatedly selected to carry out the performance indicator of overall merit classifier as training set, training obtains meeting default accuracy rate
New user's decision Tree algorithms model, for example, training obtain new user XGBoost model.The default accuracy rate can have implementation people
Member's self-setting, such as be arranged according to historical experience, or be arranged according to specific requirement.It may be configured as in the embodiment of the present invention
85%.
S305, by each old user's target histories characteristic, using decision Tree algorithms, it is default accurate that training obtains meeting
Old user's decision Tree algorithms model of rate.
In the way of the new user's decision Tree algorithms model of above-mentioned determination, by each old user's target histories characteristic,
Using decision Tree algorithms, training obtains the old user's decision Tree algorithms model for meeting default accuracy rate, for example, training is used always
Family XGBoost model, specifically repeats no more.
Pass through the training dataset and test data set of above-mentioned each sample characteristics data, it may be verified that the embodiment of the present invention obtains
New user's decision Tree algorithms model and old user's decision Tree algorithms model quality.Such as new user decision tree shown in table 2
Algorithm model and old user's decision Tree algorithms forecast result of model analytical table.
Table 2
User group | Accuracy rate precision | Recall rate recall |
New user | 76% | 68% |
Old user | 66% | 42% |
Through the embodiment of the present invention, it can be achieved that targetedly obtaining new user's decision Tree algorithms model that prediction is lost user
And old user's decision Tree algorithms model, and then new user can accurately be predicted by new user's decision Tree algorithms model
It whether is to be lost user, and can accurately predict whether old user is loss by old user's decision Tree algorithms model
User, and then the loss user of new user and the loss user of old user are targetedly arranged convenient for the later period and keep strategy, it mentions
Height keeps success rate.
A kind of loss customer analysis method of embodiment in order to better illustrate the present invention, can there is the present invention shown in Fig. 7
A kind of loss customer analysis method logical framework figure of embodiment.
Sample of users is determined first;The characteristic for extracting each sample of users carries out the characteristic of each sample of users
Data prediction obtains each new ownership goal history feature data and each old user's target histories characteristic;Pass through each new use
Family target histories characteristic, using decision Tree algorithms, training obtains the new user's decision Tree algorithms mould for meeting default accuracy rate
Type;By each old user's target histories characteristic, using decision Tree algorithms, training obtains the old user for meeting default accuracy rate
Decision Tree algorithms model.
When there is the target user to be predicted, the characteristic of the target user is extracted;With the target signature of target user
Based on data, judge target user for new user or old user;When target user is new user, by the feature of target user
Data are input to predetermined new user's decision Tree algorithms model, judge whether the target user is to be lost user;When the mesh
When mark user is old user, the characteristic of the target user is input to predetermined old user's decision Tree algorithms model,
Judge whether the target user is to be lost user.
When the target user is to be lost user, based on the characteristic of the target user, pass through high value 1 respectively
Class user's expert analysis mode table, 2 class user's expert analysis mode table of high value, 1 class user's expert analysis mode table of low value and 2 class of low value are used
Family expert analysis mode table is target user marking, and the scoring type of the highest grade form of score is determined as to the mesh of the target user
Mark is lost type.For example, target user is A, it is by the score that 1 class user's expert analysis mode table of high value obtains target user A
13;It is 32 by the score that 2 class user's expert analysis mode table of high value obtains target user A;By low value, 1 class user expert is commented
Dividing table to obtain the score of target user A is 26;It is by the score that 2 class user's expert analysis mode table of low value obtains target user A
19, then the target user A is that 2 class of high value is lost user.
Second aspect, the embodiment of the invention discloses a kind of loss customer analysis devices, as shown in Figure 8.Fig. 8 is the present invention
A kind of loss customer analysis apparatus structure schematic diagram of embodiment, device include:
Characteristic extraction module 801 for determining target user, and extracts the characteristic of target user;
It is lost user's judgment module 802, is calculated for the characteristic of target user to be input to the decision tree constructed in advance
Method model judges whether target user is to be lost user;
Target is lost determination type module 803, is used for when target user is to be lost user, with the characteristic of target user
Based on, target user is evaluated by default scoring tactics, obtains evaluation corresponding to each default scoring tactics
As a result, and the evaluation result according to corresponding to obtained default scoring tactics, determine target user target be lost type;
Tactful determining module 804 is kept, for determining that target is lost that type is corresponding keeps strategy, so that by keeping plan
Slightly keep target user.
In a kind of loss customer analysis device provided in an embodiment of the present invention, specifically, determining target user, and extract
The characteristic of target user;A more complete characteristic set has been excavated in the embodiment of the present invention for describing the row of user
For.And then the characteristic of target user is input to the decision Tree algorithms model constructed in advance, judge target user whether be
It is lost user;Using the disaggregated model of tree class, characteristic can be accurately analyzed, judges whether target user is to be lost to use
Family.When prejudging out the target user to be lost user, will further be determined by scoring tactics belonging to the loss user of anticipation
User group, finally targetedly carry out for the user group and keep strategy, promote registered users keeps success
Rate can also cut operating costs.
Optionally, in a kind of embodiment of loss customer analysis device of the invention, device further include:
Target signature data determining module carries out data prediction for the characteristic to target user, obtains target
The target signature data of user;
Target user's division module, for based on the target signature data of target user, judging according to preset rules
Target user is new user or old user;
It is lost user's judgment module 802, comprising:
First-class appraxia family judging submodule is used for when target user is new user, by the characteristic of target user
It is input to predetermined new user's decision Tree algorithms model, judges whether target user is to be lost user;
Second appraxia family judging submodule is used for when target user is old user, by the characteristic of target user
It is input to predetermined old user's decision Tree algorithms model, judges whether target user is to be lost user.
Optionally, in a kind of embodiment of loss customer analysis device of the invention, decision Tree algorithms model includes new
User's decision Tree algorithms model and old user's decision Tree algorithms model, device further include:
Sample characteristics data determining module, for choosing the history feature data of multiple users in preset time period, as
Each sample characteristics data;
Sample characteristics data division module, for each sample characteristics data to be divided into new use respectively according to preset rules
Family history feature data or old user's history feature data;
Target histories characteristic determining module, for each new user's history characteristic and each old user's history feature
Data carry out data prediction, obtain each new ownership goal history feature data and each old user's target histories characteristic;
New user's decision Tree algorithms model determining module is used for through each new ownership goal history feature data, using certainly
Plan tree algorithm, training obtain the new user's decision Tree algorithms model for meeting default accuracy rate;
Old user's decision Tree algorithms model determining module is used for through each old user's target histories characteristic, using certainly
Plan tree algorithm, training obtain the old user's decision Tree algorithms model for meeting default accuracy rate.
Optionally, in a kind of embodiment of loss customer analysis device of the invention, device further include:
Default grade form establishes module, for whether active in the recent period, frequency of use height and video playing duration with user
Length is characterized the analysis foundation of data, pre-establishes each default grade form of every kind of scoring scoring type.
The third aspect, the embodiment of the invention discloses a kind of electronic equipment, as shown in Figure 9.Fig. 9 is the embodiment of the present invention
A kind of electronic equipment structural schematic diagram, including processor 901, communication interface 902, memory 903 and communication bus 904, wherein
Processor 904, communication interface 902, memory 903 complete mutual communication by communication bus 904;
Memory 903, for storing computer program;
Processor 901 when for executing the program stored on memory 903, realizes following methods step:
It determines target user, and extracts the characteristic of target user;
The characteristic of target user is input to the decision Tree algorithms model constructed in advance, judge target user whether be
It is lost user;
When target user is to be lost user, based on the characteristic of target user, by presetting scoring tactics pair
Target user evaluates, and obtains evaluation result corresponding to each default scoring tactics, and according to obtained default scoring
Evaluation result corresponding to strategy determines that the target of target user is lost type;
Determine that target is lost the corresponding target strategy of type, so that keeping the target user by target strategy.
The communication bus 904 that above-mentioned electronic equipment is mentioned can be Peripheral Component Interconnect standard (Peripheral
Component Interconnect, abbreviation PCI) bus or expanding the industrial standard structure (Extended Industry
Standard Architecture, abbreviation EISA) bus etc..The communication bus 904 can be divided into address bus, data/address bus,
Control bus etc..Only to be indicated with a thick line in figure, it is not intended that an only bus or a type of convenient for indicating
Bus.
Communication interface 902 is for the communication between above-mentioned electronic equipment and other equipment.
Memory 903 may include random access memory (Random Access Memory, abbreviation RAM), can also be with
Including nonvolatile memory (non-volatile memory), for example, at least a magnetic disk storage.Optionally, memory
903 can also be that at least one is located remotely from the storage device of aforementioned processor 901.
Above-mentioned processor 901 can be general processor, including central processing unit (Central Processing
Unit, abbreviation CPU), network processing unit (Network Processor, abbreviation NP) etc.;It can also be digital signal processor
(Digital Signal Processing, abbreviation DSP), specific integrated circuit (Application Specific
Integrated Circuit, abbreviation ASIC), field programmable gate array (Field-Programmable Gate Array,
Abbreviation FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware components.
In a kind of electronic equipment provided in an embodiment of the present invention, specifically, determining target user, and target user is extracted
Characteristic;A more complete characteristic set has been excavated in the embodiment of the present invention for describing the behavior of user.In turn
The characteristic of target user is input to the decision Tree algorithms model constructed in advance, judges whether target user is to be lost to use
Family;Using the disaggregated model of tree class, characteristic can be accurately analyzed, judges whether target user is to be lost user.When
When prejudging out the target user to be lost user, user belonging to the loss user of anticipation will be further determined by scoring tactics
Group finally targetedly carries out for the user group and keeps strategy, promotes the success rate of keeping of registered users, also can
It cuts operating costs.
At the another aspect that the present invention is implemented, a kind of computer readable storage medium is additionally provided, it is described computer-readable
It is stored with instruction in storage medium, when run on a computer, realizes any described in above-mentioned loss customer analysis method
Method and step.
In a kind of computer readable storage medium provided in an embodiment of the present invention, specifically, determining target user, and mention
Take the characteristic of target user;A more complete characteristic set has been excavated in the embodiment of the present invention for describing user's
Behavior.And then the characteristic of target user is input to the decision Tree algorithms model constructed in advance, whether judge target user
To be lost user;Using the disaggregated model of tree class, characteristic can be accurately analyzed, judges whether target user is loss
User.When prejudging out the target user to be lost user, the loss user institute of anticipation will be further determined by scoring tactics
The user group of category finally targetedly carries out for the user group and keeps strategy, promotes keeping into for registered users
Power can also cut operating costs.
At the another aspect that the present invention is implemented, the embodiment of the invention also provides a kind of, and the computer program comprising instruction is produced
Product realize any method and step in above-mentioned loss customer analysis method when run on a computer.
In a kind of computer program product comprising instruction provided in an embodiment of the present invention, specifically, determining that target is used
Family, and extract the characteristic of target user;A more complete characteristic set has been excavated in the embodiment of the present invention for retouching
State the behavior of user.And then the characteristic of target user is input to the decision Tree algorithms model constructed in advance, judge target
Whether user is to be lost user;Using the disaggregated model of tree class, characteristic can be accurately analyzed, judges that target user is
No is to be lost user.When prejudging out the target user to be lost user, the stream of anticipation will be further determined by scoring tactics
User group belonging to appraxia family finally targetedly carries out for the user group and keeps strategy, promotes registered users
Keep success rate, can also cut operating costs.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or any combination thereof real
It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.The computer program
Product includes one or more computer instructions.When loading on computers and executing the computer program instructions, all or
It partly generates according to process or function described in the embodiment of the present invention.The computer can be general purpose computer, dedicated meter
Calculation machine, computer network or other programmable devices.The computer instruction can store in computer readable storage medium
In, or from a computer readable storage medium to the transmission of another computer readable storage medium, for example, the computer
Instruction can pass through wired (such as coaxial cable, optical fiber, number from a web-site, computer, server or data center
User's line (DSL)) or wireless (such as infrared, wireless, microwave etc.) mode to another web-site, computer, server or
Data center is transmitted.The computer readable storage medium can be any usable medium that computer can access or
It is comprising data storage devices such as one or more usable mediums integrated server, data centers.The usable medium can be with
It is magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or semiconductor medium (such as solid state hard disk
Solid State Disk (SSD)) etc..
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
Each embodiment in this specification is all made of relevant mode and describes, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for device,
For electronic equipment and storage medium embodiment, since it is substantially similar to the method embodiment, so being described relatively simple, phase
Place is closed to illustrate referring to the part of embodiment of the method.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all
Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention
It is interior.
Claims (10)
1. a kind of loss customer analysis method, which is characterized in that the described method includes:
It determines target user, and extracts the characteristic of the target user;
The characteristic of the target user is input to preset decision Tree algorithms model, judge the target user whether be
It is lost user;
When the target user is to be lost user, based on the characteristic of the target user, pass through default scoring plan
Slightly the target user is evaluated, obtains evaluation result corresponding to each default scoring tactics, and according to gained
Evaluation result corresponding to the default scoring tactics arrived determines that the target of the target user is lost type;
Determine that the target is lost that type is corresponding to keep strategy, so that keeping strategy by described and keeping the target user.
2. the method according to claim 1, wherein in the determining target user, and extracting the target and using
After the characteristic at family, the method also includes:
Data prediction is carried out to the characteristic of the target user, obtains characteristic needed for analyzing the target user
According to as target signature data;
According to preset rules, based on the target signature data of the target user, judge the target user for new user
Or old user;
The characteristic by the target user is input to preset decision Tree algorithms model, judges that the target user is
No is to be lost user, comprising:
When the target user is new user, the characteristic of the target user is input to preset new user decision tree
Algorithm model judges whether the target user is to be lost user;
When the target user is old user, the characteristic of the target user is input to preset old user decision tree
Algorithm model judges whether the target user is to be lost user.
3. according to the method described in claim 2, it is characterized in that, construct the preset decision Tree algorithms model the step of,
Include:
The history feature data for choosing multiple users in preset time period, as each sample characteristics data;
According to preset rules, each sample characteristics data are divided into new user's history characteristic or old user's history respectively
Characteristic;
To each new user's history characteristic and each old user's history feature data, data prediction is carried out, is obtained
Characteristic needed for analyzing each new user, as feature needed for new ownership goal history feature data and each old user of analysis
Data, as old user's target histories characteristic;
By each new ownership goal history feature data, using decision Tree algorithms, training obtains meeting default accuracy rate
New user's decision Tree algorithms model;
By each old user's target histories characteristic, using decision Tree algorithms, it is described default accurate that training obtains meeting
Old user's decision Tree algorithms model of rate.
4. whether being enlivened, being made in the recent period with user the method according to claim 1, wherein the scoring tactics are
It is established with the analysis foundation that frequency height and video playing duration length are characterized data.
5. a kind of loss customer analysis device, which is characterized in that described device includes:
Characteristic extraction module for determining target user, and extracts the characteristic of the target user;
It is lost user's judgment module, for the characteristic of the target user to be input to preset decision Tree algorithms model,
Judge whether the target user is to be lost user;
Target is lost determination type module, is used for when the target user is to be lost user, with the feature of the target user
Based on data, the target user is evaluated by default scoring tactics, obtains each default scoring tactics institute
Corresponding evaluation result, and the evaluation result according to corresponding to the obtained default scoring tactics determine that the target is used
The target at family is lost type;
Tactful determining module is kept, for determining that the target is lost that type is corresponding to keep strategy, so that keeping by described
Strategy keeps the target user.
6. device according to claim 5, which is characterized in that described device further include:
Target signature data determining module carries out data prediction for the characteristic to the target user, is analyzed
Characteristic needed for the target user, as target signature data;
Target user's division module, for based on the target signature data of the target user, judging according to preset rules
The target user is new user or old user;
Loss user's judgment module, comprising:
First-class appraxia family judging submodule is used for when the target user is new user, by the feature of the target user
Data are input to preset new user's decision Tree algorithms model, judge whether the target user is to be lost user;
Second appraxia family judging submodule is used for when the target user is old user, by the feature of the target user
Data are input to preset old user's decision Tree algorithms model, judge whether the target user is to be lost user.
7. device according to claim 6, which is characterized in that described device further include:
Sample characteristics data determining module, for choosing the history feature data of multiple users in preset time period, as various kinds
Eigen data;
Sample characteristics data division module, for each sample characteristics data to be divided into new use respectively according to preset rules
Family history feature data or old user's history feature data;
Target histories characteristic determining module, for each new user's history characteristic and each old user's history
Characteristic carries out data prediction, characteristic needed for analyzing each new user, as new ownership goal history feature data
And characteristic needed for each old user of analysis, as old user's target histories characteristic;
New user's decision Tree algorithms model determining module is used for through each new ownership goal history feature data, using certainly
Plan tree algorithm, training obtain the new user's decision Tree algorithms model for meeting default accuracy rate;
Old user's decision Tree algorithms model determining module is used for through each old user's target histories characteristic, using certainly
Plan tree algorithm, training obtain the old user's decision Tree algorithms model for meeting the default accuracy rate.
8. device according to claim 5, which is characterized in that whether the scoring tactics are to be enlivened, made in the recent period with user
It is established with the analysis foundation that frequency height and video playing duration length are characterized data.
9. a kind of electronic equipment, which is characterized in that including processor, communication interface, memory and communication bus, wherein described
Processor, the communication interface, the memory complete mutual communication by the communication bus;
The memory, for storing computer program;
The processor when for executing the program stored on the memory, realizes any side claim 1-4
Method step.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer in the computer readable storage medium
Program realizes claim 1-4 any method and step when the computer program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910473357.4A CN110222975A (en) | 2019-05-31 | 2019-05-31 | A kind of loss customer analysis method, apparatus, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910473357.4A CN110222975A (en) | 2019-05-31 | 2019-05-31 | A kind of loss customer analysis method, apparatus, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110222975A true CN110222975A (en) | 2019-09-10 |
Family
ID=67819383
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910473357.4A Pending CN110222975A (en) | 2019-05-31 | 2019-05-31 | A kind of loss customer analysis method, apparatus, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110222975A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178973A (en) * | 2019-12-31 | 2020-05-19 | 北京每日优鲜电子商务有限公司 | Loss user identification recognition method and device, computer equipment and storage medium |
CN111325006A (en) * | 2020-03-17 | 2020-06-23 | 北京百度网讯科技有限公司 | Information interaction method and device, electronic equipment and storage medium |
CN111401969A (en) * | 2020-04-02 | 2020-07-10 | 北京达佳互联信息技术有限公司 | Method, device, server and storage medium for improving user retention rate |
CN112016790A (en) * | 2020-07-15 | 2020-12-01 | 北京淇瑀信息科技有限公司 | User policy distribution method and device and electronic equipment |
CN112862527A (en) * | 2021-02-04 | 2021-05-28 | 北京嘀嘀无限科技发展有限公司 | User type determination method, device, equipment and storage medium |
CN113191701A (en) * | 2021-07-02 | 2021-07-30 | 深圳市知酷信息技术有限公司 | User management system based on internet document sharing space |
CN113256044A (en) * | 2020-02-13 | 2021-08-13 | ***通信集团广东有限公司 | Strategy determination method and device and electronic equipment |
CN113283922A (en) * | 2020-02-20 | 2021-08-20 | 百度在线网络技术(北京)有限公司 | Lost user saving method, device, equipment and medium |
CN114191824A (en) * | 2021-11-30 | 2022-03-18 | 上海乐畅信息技术有限公司 | Method and device for sending friend recall information to user |
CN114549071A (en) * | 2022-02-18 | 2022-05-27 | 上海钧正网络科技有限公司 | Marketing strategy determination method and device, computer equipment and storage medium |
CN114584601A (en) * | 2022-01-26 | 2022-06-03 | 上海钧正网络科技有限公司 | User loss identification and intervention method, system, terminal and medium |
CN114742144A (en) * | 2022-04-02 | 2022-07-12 | 中国建设银行股份有限公司 | Method and device for determining object attrition probability, computer equipment and storage medium |
CN115086705A (en) * | 2021-03-12 | 2022-09-20 | 北京字跳网络技术有限公司 | Resource preloading method, device, equipment and storage medium |
CN116051154A (en) * | 2023-03-06 | 2023-05-02 | 吉林省国迅广告有限公司 | Media terminal data analysis management system |
CN116385080A (en) * | 2023-04-17 | 2023-07-04 | 云洞(上海)科技股份有限公司 | Mobile internet user data statistics popularization system based on artificial intelligence |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070156673A1 (en) * | 2005-12-30 | 2007-07-05 | Accenture S.P.A. | Churn prediction and management system |
CN106022800A (en) * | 2016-05-16 | 2016-10-12 | 北京百分点信息科技有限公司 | User feature data processing method and device |
CN106203679A (en) * | 2016-06-27 | 2016-12-07 | 武汉斗鱼网络科技有限公司 | A kind of customer loss Forecasting Methodology and system |
CN107833063A (en) * | 2017-11-30 | 2018-03-23 | 上海科瓴医疗科技有限公司 | Pharmacy member is lost in early warning and intelligent interfering system and method |
CN108665321A (en) * | 2018-05-18 | 2018-10-16 | 广州虎牙信息科技有限公司 | High viscosity customer loss prediction technique, device and computer readable storage medium |
CN109285011A (en) * | 2018-09-12 | 2019-01-29 | 田雪娇 | A kind of smart client return visit method, device and equipment |
CN109636443A (en) * | 2018-11-17 | 2019-04-16 | 南京中数媒介研究有限公司 | The deep learning method and device of customer churn prediction |
CN109685255A (en) * | 2018-12-13 | 2019-04-26 | 麒麟合盛网络技术股份有限公司 | A kind of method and apparatus for predicting customer churn |
-
2019
- 2019-05-31 CN CN201910473357.4A patent/CN110222975A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070156673A1 (en) * | 2005-12-30 | 2007-07-05 | Accenture S.P.A. | Churn prediction and management system |
CN106022800A (en) * | 2016-05-16 | 2016-10-12 | 北京百分点信息科技有限公司 | User feature data processing method and device |
CN106203679A (en) * | 2016-06-27 | 2016-12-07 | 武汉斗鱼网络科技有限公司 | A kind of customer loss Forecasting Methodology and system |
CN107833063A (en) * | 2017-11-30 | 2018-03-23 | 上海科瓴医疗科技有限公司 | Pharmacy member is lost in early warning and intelligent interfering system and method |
CN108665321A (en) * | 2018-05-18 | 2018-10-16 | 广州虎牙信息科技有限公司 | High viscosity customer loss prediction technique, device and computer readable storage medium |
CN109285011A (en) * | 2018-09-12 | 2019-01-29 | 田雪娇 | A kind of smart client return visit method, device and equipment |
CN109636443A (en) * | 2018-11-17 | 2019-04-16 | 南京中数媒介研究有限公司 | The deep learning method and device of customer churn prediction |
CN109685255A (en) * | 2018-12-13 | 2019-04-26 | 麒麟合盛网络技术股份有限公司 | A kind of method and apparatus for predicting customer churn |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178973A (en) * | 2019-12-31 | 2020-05-19 | 北京每日优鲜电子商务有限公司 | Loss user identification recognition method and device, computer equipment and storage medium |
CN113256044B (en) * | 2020-02-13 | 2023-08-15 | ***通信集团广东有限公司 | Policy determination method and device and electronic equipment |
CN113256044A (en) * | 2020-02-13 | 2021-08-13 | ***通信集团广东有限公司 | Strategy determination method and device and electronic equipment |
CN113283922A (en) * | 2020-02-20 | 2021-08-20 | 百度在线网络技术(北京)有限公司 | Lost user saving method, device, equipment and medium |
CN111325006A (en) * | 2020-03-17 | 2020-06-23 | 北京百度网讯科技有限公司 | Information interaction method and device, electronic equipment and storage medium |
CN111325006B (en) * | 2020-03-17 | 2023-05-05 | 北京百度网讯科技有限公司 | Information interaction method and device, electronic equipment and storage medium |
CN111401969A (en) * | 2020-04-02 | 2020-07-10 | 北京达佳互联信息技术有限公司 | Method, device, server and storage medium for improving user retention rate |
CN111401969B (en) * | 2020-04-02 | 2024-02-20 | 北京达佳互联信息技术有限公司 | Method, device, server and storage medium for improving user retention |
CN112016790A (en) * | 2020-07-15 | 2020-12-01 | 北京淇瑀信息科技有限公司 | User policy distribution method and device and electronic equipment |
CN112016790B (en) * | 2020-07-15 | 2024-04-26 | 北京淇瑀信息科技有限公司 | User policy allocation method and device and electronic equipment |
CN112862527A (en) * | 2021-02-04 | 2021-05-28 | 北京嘀嘀无限科技发展有限公司 | User type determination method, device, equipment and storage medium |
CN115086705A (en) * | 2021-03-12 | 2022-09-20 | 北京字跳网络技术有限公司 | Resource preloading method, device, equipment and storage medium |
CN113191701A (en) * | 2021-07-02 | 2021-07-30 | 深圳市知酷信息技术有限公司 | User management system based on internet document sharing space |
CN114191824A (en) * | 2021-11-30 | 2022-03-18 | 上海乐畅信息技术有限公司 | Method and device for sending friend recall information to user |
CN114584601A (en) * | 2022-01-26 | 2022-06-03 | 上海钧正网络科技有限公司 | User loss identification and intervention method, system, terminal and medium |
CN114549071A (en) * | 2022-02-18 | 2022-05-27 | 上海钧正网络科技有限公司 | Marketing strategy determination method and device, computer equipment and storage medium |
CN114742144A (en) * | 2022-04-02 | 2022-07-12 | 中国建设银行股份有限公司 | Method and device for determining object attrition probability, computer equipment and storage medium |
CN116051154A (en) * | 2023-03-06 | 2023-05-02 | 吉林省国迅广告有限公司 | Media terminal data analysis management system |
CN116385080A (en) * | 2023-04-17 | 2023-07-04 | 云洞(上海)科技股份有限公司 | Mobile internet user data statistics popularization system based on artificial intelligence |
CN116385080B (en) * | 2023-04-17 | 2024-01-26 | 云洞(上海)科技股份有限公司 | Mobile internet user data statistics popularization system based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110222975A (en) | A kind of loss customer analysis method, apparatus, electronic equipment and storage medium | |
Lepschy et al. | Success factors in football: an analysis of the German Bundesliga | |
US20190163690A1 (en) | Keyword extraction method, apparatus and server | |
CN106503014B (en) | Real-time information recommendation method, device and system | |
CN105045831B (en) | A kind of information push method and device | |
CN106372249B (en) | A kind of clicking rate predictor method, device and electronic equipment | |
Barros et al. | Coaching for survival: The hazards of head coach careers in the German ‘Bundesliga’ | |
CN109948121A (en) | Article similarity method for digging, system, equipment and storage medium | |
CN107689008A (en) | A kind of user insures the method and device of behavior prediction | |
CN108304853B (en) | Game correlation obtaining method and device, storage medium and electronic device | |
CN106446078A (en) | Information recommendation method and recommendation apparatus | |
CN113688167A (en) | Deep interest capture model construction method and device based on deep interest network | |
CN109582875A (en) | A kind of personalized recommendation method and system of online medical education resource | |
CN105225135B (en) | Potential customer identification method and device | |
CN104077417A (en) | Figure tag recommendation method and system in social network | |
CN108345601A (en) | Search result ordering method and device | |
CN102110098A (en) | Method and system for recommending network information | |
Parmar et al. | Team performance indicators that predict match outcome and points difference in professional rugby league | |
CN107239564A (en) | A kind of text label based on supervision topic model recommends method | |
CN109961311A (en) | Lead referral method, apparatus calculates equipment and storage medium | |
Wunderlich et al. | A big data analysis of Twitter data during premier league matches: do tweets contain information valuable for in-play forecasting of goals in football? | |
Hubáček et al. | Forty years of score-based soccer match outcome prediction: an experimental review | |
CN106326244A (en) | Method and apparatus for determining user interest in web-based content | |
CN108009194A (en) | A kind of books method for pushing, electronic equipment, storage medium and device | |
CN108664605B (en) | Model evaluation method and system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190910 |
|
RJ01 | Rejection of invention patent application after publication |