CN107203518A - Method, system and device, the electronic equipment of on-line system personalized recommendation - Google Patents
Method, system and device, the electronic equipment of on-line system personalized recommendation Download PDFInfo
- Publication number
- CN107203518A CN107203518A CN201610149438.5A CN201610149438A CN107203518A CN 107203518 A CN107203518 A CN 107203518A CN 201610149438 A CN201610149438 A CN 201610149438A CN 107203518 A CN107203518 A CN 107203518A
- Authority
- CN
- China
- Prior art keywords
- user
- personalized recommendation
- model
- access
- model training
- 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
- 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
- G06Q30/0282—Rating or review of business operators or products
-
- 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
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Theoretical Computer Science (AREA)
- Development Economics (AREA)
- Databases & Information Systems (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
This application discloses a kind of method, device and the electronic equipment of on-line system personalized recommendation, and a kind of system of on-line system personalized recommendation.The personalized recommendation information that the method for wherein described on-line system personalized recommendation makes feedback operation to user characteristic data, personalized recommendation information and to the user is collected in real time, and form the model training entry that Adds User, the training sample set of entry formation is trained by accumulating the user model, train the user model, and after new user model is formed, immediately available as active user's model, for providing a user personalized recommendation information.The method provided using the application, while personalized recommendation information is provided a user, is carried out information search and forms user model training entry；This processing mode, can effectively reduce information and collect the amount of calculation of generation, and reduce the memory space of storage average information；Therefore, the method that the application is provided can effectively reduce resource consumption.
Description
Technical field
The application is related to field of computer technology, and in particular to a kind of method of on-line system personalized recommendation；
Corresponding to the above method, the application is related to a kind of device and electronic equipment of online personalized recommendation simultaneously,
And a kind of system of on-line system personalized recommendation.
Background technology
The fast development of Internet technology makes the mankind enter epoch of information explosion.It is in while magnanimity information
Existing, on the one hand making information acquisition, person is difficult therefrom to find oneself part interested, is on the other hand also caused big
The information that the few people of amount makes inquiries can not be obtained by general user, and case above seriously hinders the abundant of information utility
Play.
To solve the above problems, occurring in that personalized recommendation system at present.Personalized recommendation system is by setting up
Binary crelation between user and information, each user is excavated using existing selection course or similarity relationships
Potential customized information interested, and then personalized recommendation is carried out, make to have between information and information user
There is higher matching degree.
Personalized recommendation system under prior art, is realized in the following way.
A large amount of behaviors for collecting user in advance, extract sample data first, according to these sample datas, lead to
The method for crossing machine learning, training obtain user model, online according to the user characteristic data of each user with
And above-mentioned user model, it is that user recommends customized information, in online trading system, the personalized letter
Breath generally comprises commodity, shop or brand etc. corresponding to the entity of personalized recommendation；For example, Amazon
Website, according to user characteristic data recommended book, is exactly the example of personalized recommendation entity.
, it is necessary to collect mass data by different system under above-mentioned prior art, and it is original by what is collected
Data, are aggregated into offline platform.For example, user is exposed into daily record, click logs, fetched data, shopping
The data of multiple different applications such as garage is, favorites data, are aggregated into offline platform.Then, according to this
A little initial data calculate the feature and characteristic of all users, further according to the user recorded in these data
Feedback behavior, such as：Click, access, thumb up, collection, predetermined, purchase etc., mark is carried out to sample,
Finally model training is carried out in offline big data platform.Training is finished, and the user model of acquisition, which is reached the standard grade, to be made
With.
There is open defect in above-mentioned prior art, subject matter is, it is necessary to flowed back data from multiple platforms,
Then collect, calculate, this process needs to expend more computing resource and storage resource；Further, since
The efficiency and time point of each platform backflow data differ, and cause model training to need to wait the long period, make instruction
Practice the poor real of model.The data volume for once needing to accumulate is additionally, since than larger, can only typically be used
A part of data record collected, the situation of many data reflections can not reflect in user model, it is impossible to
Really implement big data application.
The content of the invention
The application provides a kind of method of on-line system personalized recommendation, to solve resource consumption under prior art
Excessively, the problem of poor real, and big data application is really implemented.The application also provides a kind of online individual character
Change commending system, and on-line system personalized recommendation device；And one kind realizes that on-line system personalization is pushed away
The electronic equipment recommended.
The application provides a kind of method of on-line system personalized recommendation, including：
The access request for accessing user is received, and extracts the user characteristic data of the access user；
According to the user characteristic data of the access user, and there is provided personalized recommendation for active user's model
Information；
The user characteristic data for collecting the access user in real time, the individual character provided for the access user
Change recommendation information, the access user make the personalized recommendation information of feedback operation and done it is anti-
Feedback operation, and form the model training entry that Adds User；
The model training entry that Adds User is added into training sample set；
User model training is carried out with current training sample set；
After the completion of the user model training, the user model of the renewal obtained is regard as active user's model.
Optionally, the user characteristic data at least includes：User Identity, and including following any use
At least one in the characteristic of family：Sex, age, transaction record.
Optionally, the personalized recommendation information includes at least one of following personalized recommendation entity：Commodity,
Shop, brand.
Optionally, the access user, which makes feedback operation, includes one of following operation：Click on, access, point
Praise, collect, make a reservation for, purchase.
Optionally, the one or two kinds of of data below is also included in the training data entry：Obtain the note
The time point of record, resource-niche information.
Optionally, the model training entry that Adds User that will build up on is added after training sample set, described
Before the step of carrying out user model training with current training sample set, following step is performed：
Whether the model training sample that Added User described in judging reaches predetermined threshold value；If so, then entering next
Step.
Optionally, the user model training uses machine learning method.
Optionally, the machine learning method is using logistic regression method or gradient lifting traditional decision-tree.
Optionally, the model training entry that Adds User is recorded using log mode.
Accordingly, the application also provides a kind of device of on-line system personalized recommendation, including：
User characteristic data extraction unit, the access request of user is accessed for receiving, and extracts the access
The user characteristic data of user；
Personalized recommendation information provider unit, for the user characteristic data according to the access user, and
There is provided personalized recommendation information for active user's model；
User model training entry formation unit, for collect in real time it is described access user user characteristic data,
The institute of feedback operation is made for the personalized recommendation information of the access user offer, the access user
Personalized recommendation information and the feedback operation done are stated, and forms the model training entry that Adds User；
Training sample set collects unit, for the model training entry that Adds User to be added into training sample set；
User model training unit, for carrying out user model training with current training sample set；
User model updating block, for after the completion of user model training, by the renewal obtained
User model is used as active user's model.
Accordingly, the application also provides a kind of electronic equipment, including：
Display；
Processor；
Memory, the program for storing the method for realizing on-line system personalized recommendation, equipment is powered simultaneously
After the program for the method for running the on-line system personalized recommendation, following step is performed：
The access request for accessing user is received, and extracts the user characteristic data of the access user；
According to the user characteristic data of the access user, and there is provided personalized recommendation for active user's model
Information；
The user characteristic data for collecting the access user in real time, the individual character provided for the access user
Change recommendation information, the access user make the personalized recommendation information of feedback operation and done it is anti-
Operation is presented, and the above of once access-recommendation process formation to access user is combined as one and is increased newly
User model training entry；
The newly-increased user model training entry is added into training sample set；
User model training is carried out with current training sample set；
After the completion of the user model training, the user model of the renewal obtained is regard as active user's model.
Accordingly, the application also provides a kind of system of online personalized recommendation, including：Online subsystem,
Offline subsystem；
The online subsystem, for receiving the access request of user, and extracts user characteristic data, and root
According to the user characteristic data and active user's model extracted, individual character is provided to the user for proposing access request
Change recommended entity；And, the user characteristic data of the access user is collected in real time, be the access user
The personalized recommendation information that there is provided, the access user to the feedback information of the personalized recommendation information,
Formation, which Adds User, model training entry and to be sent；And, receive the renewal that the offline subsystem is provided
User model；
The offline subsystem, receives the model training entry that Adds User that the online subsystem is sent, and
The model training entry that Adds User is added into current training sample set；Using the current training sample set
User model training is carried out, the user model for training the renewal for completing to obtain is sent out.
Compared with prior art, the method for the on-line system personalized recommendation that the application is provided, to user characteristics
Data, personalized recommendation information and the personalized recommendation information progress that feedback operation is made to the user
Collect in real time, and form the model training entry that Adds User, entry shape is trained by accumulating the user model
Into training sample set, the user model is trained, and after new user model is formed, immediately available as working as
Preceding user model, for providing a user personalized recommendation information.
The method provided using the application, while personalized recommendation information is provided a user, enters row information
Collect and form user model training entry；This processing mode, can effectively reduce information and collect generation
Amount of calculation, and reduce the memory space of storage average information；Therefore, the method that the application is provided can be with
Effectively reduce resource consumption.
In the method that the application is provided, user model training entry can be used for carrying out user model instruction at any time
Practice, and the user model newly formed can be used when providing a user personalized recommendation information at once.This
Sample, can be adjusted, and be readily used for providing a user according to the data newly collected to user model at any time
Personalized recommendation information；Therefore, the method that the application is provided can be in time according to the data collected to user
Model is adjusted, and has more high real-time than prior art.In addition, the method that the application is provided, moreover it is possible to
Collected total data is enough made full use of, big data application is effectively realized.
Brief description of the drawings
Fig. 1 is a kind of method flow diagram for on-line system personalized recommendation that the application first embodiment is provided；
Fig. 2 is a kind of unit frame of the device for on-line system personalized recommendation that the application second embodiment is provided
Figure；
Fig. 3 is a kind of system schematic for on-line system personalized recommendation that the application fourth embodiment is provided.
Embodiment
Many details are elaborated in the following description to fully understand the application.But, this Shen
It is able to please be implemented with being much different from other manner described here, those skilled in the art can not disobey
Similar popularization is done in the case of back of the body the application intension, therefore, the application is not by following public specific implementation
Limitation.
This application provides a kind of method and apparatus of on-line system personalized recommendation, and a kind of online individual character
Change the system recommended；And the electronic equipment of on-line system personalized recommendation, in the following embodiments one by one
It is described in detail.
For the ease of understanding the technical scheme of the application, first to proposing the background of the application and the skill of the application
Art scheme is briefly described.
Reference picture 1, a kind of side of the on-line system personalized recommendation provided it illustrates the application first embodiment
Method process chart.
In this embodiment, system is according to user characteristic data, and there is provided personalization for active user's model
Recommended entity；Also, this described recommending related user characteristic data, providing the user can be recorded
Property recommendation information, and user makes the situation of feedback operation, and as training new user model
Data.
For example：When user is buying a mobile phone, system will be recommended to be adapted to use according to user characteristic data
The earphone that family needs, the commodity such as cell-phone cover；User makes purchase operation to certain earphone；Said process may
A plurality of user model training entry is formed, wherein the model training entry that Adds User, including following data：
The user characteristic data of the user, the earphone recommended to user and the user will make purchase behaviour to earphone
The mark record of work, this user model training entry is used as positive sample；The other one model instruction that Adds User
Practicing entry includes following data：The user characteristic data of the user, the cell-phone cover recommended to user, and should
User does not make the mark record of operation to cell-phone cover, and this user model training entry can be used as negative sample
Use.
Illustrated below in conjunction with a kind of personalized recommendation system method that Fig. 1 is provided the present embodiment, and
Each step to this method is illustrated.
Step S101, receives the access request for accessing user, and extract the user characteristics number of the access user
According to.
So-called access user, refer mainly to by internet or mobile network online access particular station, webpage,
The user of server.
The user characteristic data for accessing user, refers to that reflection accesses each side characteristic information of user in itself
Data.These data can be obtained by different channels.
First, access user it is general by browser or APP using intermediary is used as, realize to particular station,
The access of webpage, server.These particular stations, webpage or server is logged in generally require by account
Log in, the related characteristic of user can be recorded by corresponding to these accounts.
Secondly, user characteristic data can also need not be used by user's access mode, customer access area domain etc.
The mode of family logon account is obtained.For example, the access mode feature for accessing user includes：Internet is visited
Ask or mobile Internet is accessed；During mobile Internet is accessed, it can also determine whether that the access used is whole
The brand at end；The customer access area domain, i.e. LBS information, i.e., the region where when user conducts interviews,
These regions can be divided into different type, such as colleges and universities, business office place according to specified place；Such as city,
Small towns, each type reflects that user may influence the feature of its feedback operation behavior with different.
The user characteristic data, is divided from acquisition pattern, including the data that user oneself provides, Yi Jigen
The related data of access behavior acquisition is logged according to user.
The data that the user oneself provides, including the User Identity that user is provided when logging in, and note
The sex that there is provided during volume, the information such as age；It can also include what user was associated in using website, server
The information such as mailbox, Bank Account Number.
It is described according to user log in access behavior acquisition related data, including the data directly obtained and indirectly
The data of acquisition.
The data directly obtained include：User of the user to access, the acquisition of lower one-state of shopping website
Record, user's purchaser record and other various user behavior information are accessed, these information are to user's history
The direct record of behavior.
The secondhand data, the reflection user mainly summed up from the above-mentioned data directly obtained
The data of feature, for example, the books bought according to user, educational level, reading field to user are made
Classification etc..The Brand bought according to user, judgement that the purchasing power level to user is made etc..
Under different application scenarios, the user characteristic data specifically can include different detailed programs, but
No matter which kind of situation, at least including User Identity (ID), User Identity directly can pass through user
The information acquisition provided when logging in, it is possible to obtained as further inquiry user other information record according to
According to；For example, the sex of user, age, and transaction record (under the scene of shopping at network) etc., these
Information using User Identity as major key as it was previously stated, be typically stored in the record sheet of database.
The tables of data or database for recording user characteristic data both can be on long-range servers, it is also possible in visitor
Family end.
The various channels and acquisition pattern described above for obtaining user characteristic data, in a particular embodiment, root
According to different demands, it is necessary to collect required user characteristic data, these user characteristicses as the case may be
Data should be that useful data are trained to user model.
In general, the user characteristic data at least includes User Identity, and including following any user
At least one in characteristic：Sex, age, transaction record；Certainly, user characteristic data completely can be with
Include other possible related data.With the development and the progress of data mining technology of big data technology,
Increasing data dependence is found, and has the data model become better and better to reflect various data
To finally desirable to provide personalized recommendation information effect, therefore, it can for user model train use
The type of family characteristic also can be more and more.
Access request alleged by this step, the access that can include to specific website, webpage is browsed, can also
Refer to search of the user in website to commodity etc..
Step S102, according to it is described access user user characteristic data, and active user's model there is provided
Personalized recommendation information.
By abovementioned steps S101, the user characteristic data for accessing user is obtained；These user characteristic datas
The foundation for providing the user personalized recommendation information can be used as；The specific personalized recommendation information that obtains needs to lead to
Cross using active user's model, using the user characteristic data as foundation, bring active user's model into, derive
Go out the personalized recommendation information.
Active user's model, is that the user of each user collected before being accessed according to the access user is special
Data are levied, the user model being being currently used of acquisition is trained in a predetermined manner.Active user's mould
The effect of type is provided it after required specific user characteristic data, and active user's model being capable of root
The user personality reflected according to user characteristic data, corresponding personalized recommendation information is provided to the user.
In the present embodiment, by continuous gather data, new training data can be constantly accumulated, these instructions
Practice data and can be used for the training to the user model, so as to obtain the user model of renewal.This step makes
With the word of active user's model one, wherein " current " is explanation user model be in the present embodiment continuous amendment,
Develop, the user model that this step is used is currently valid user model, but after a while, the use
Family model just may be varied from due to new training process.
The personalized recommendation information, refers to that active user's model is special according to the user of specific access user
Levy the information recommended to specific access user that data are extrapolated.It is described to recommend, refer mainly to use in the access
The access showing interface at family, including vision, the sense of hearing or the displaying of other possible modes；The personalization,
Its implication refers to that the user characteristics with the user matches.
The personalized recommendation information can correspond to different types of personalized recommendation entity as the case may be,
In this embodiment it is assumed that scene be electric business sale scene, then the personalized recommendation information can be as
The personalized recommendation entity of Types Below：Commodity, shop or brand.
For example, certain position, which accesses user, accesses internet book store, the use reflected according to its user characteristic data
The interest range at family, can be the related books commodity of its recommendation；Certain accesses user and accesses Taobao website, searches
Rope notebook, the price endurance or Brand Preference of the user reflected according to its user characteristic data,
Its user characteristics can be met for its recommendation and the shop on net of notebook is sold；Certain accesses user and inquired about
Clothes are bought, sex, age and the purchasing power of the user reflected according to its user characteristics can be with
The apparel brand for meeting its sex, age and hierarchy of consumption is provided for its recommendation.
Active user's model can include one or more submodel, be each responsible for be from different perspectives
Access the personalized recommendation that user carries out different aspect information；And can be finally in the exhibition of same access interface
It is existing, it is of course also possible to which the concrete condition accessed according to user shows the son using specific active user's model
Model, for example, after a user being already registered for is by Account Logon, at once according to its user characteristics number
According to showing commodity, shop or the brand that the emerging user may be concerned about after its last login to it；
Or, user search books when, at once using be responsible for recommended book active user's model submodel according to
User characteristic data is that user recommends appropriate books.The specific user characteristic data that these submodels are used,
May be a specific part of all user characteristic datas of access user respectively.These different submodules
Type can also be considered as the different function units of same active user's model.
Step S103, collects the user characteristic data of the access user, is provided for the access user in real time
The personalized recommendation information, the access user make the personalized recommendation information of feedback operation with
And the feedback operation done, and form the model training entry that Adds User.
This step is used to collect user's access, provides a user personalized recommendation information and user to personalization
Recommendation information feed back all related contents of this complete access-recommendation process formation, including：Individual character
Change the original foundation recommended, i.e. user characteristic data；The personalized recommendation information provided for the access user；
To the feedback of personalized recommendation, i.e., described access user which kind of personalized recommendation information has been made feedback and
Which kind of feedback specifically made.Most these contents combine to form data record at last.Once access-recommendation process shape
Into the data record can be recorded using different concrete forms.For example it is possible to record complete for one
Packet, be sent to backstage, dependency number therefrom extracted according to the demand of model training by background server
According to a plurality of model training entry that Adds User of formation.It is of course also possible to directly form the model training that Adds User
Entry is simultaneously sent.In a word, it is necessary to collect and parse by rights one the need for being trained according to user model
The data record of secondary access-recommendation process formation.
The model training entry that Adds User, the training material trained as follow-up user model.According to
Need, the information collected can further include the time point for obtaining the record and resource-niche information,
The information that the time point for obtaining the record, i.e. record form the newly-increased user model training entry is formed
Time；The resource-niche information, i.e. position where recording-related information, for example, some specific individual character
Change recommendation information in the top of browser either bottom or sidebar, etc., these information are for user's
Feedback behavior also has a major impact.
This step is realized, it is necessary to for accessing implant procedure, the program in the browser that uses of user or app
The feedback of all personalized recommendation information provided a user and user to personalized recommendation information is collected at any time
Operation, these information are corresponding with the user characteristic data obtained previously as personalized recommendation information,
And the server side of training user model is sent to by network.Specifically used implant procedure can basis
Varying environment uses different types, for example, using Javascript shell scripts in a browser to by clear
The various information of device of looking at offer and the various operations occurred by browser interface are recorded.
It is described access user feedback operation, including access user the personalized recommendation information is made it is each
The feedback of the mode of kind, for example：Click on, access, thumb up, collect, make a reservation for, the operation such as purchase.These behaviour
Reacted and accessed user to the difference of the degree of attentiveness of different personalized recommendation information, these differences for
Adjust the customized information recommended to user significant.
Record it is described access user make feedback operation the personalized recommendation information method, can be
Specific identifier is added in these information, i.e. plus expression to it on the record of these personalized recommendation information
It there occurs the mark of feedback operation, commonly referred to as mark；For example, correspondence is pushed away per personalization in record form
Recommend information and be provided with the field for identifying whether that feedback operation is there occurs thereon, this field can also directly be remembered
What the feedback operation that record occurs is, by this recording mode, just in the same of record personalized recommendation information
When to access user feedback recorded.
The above of the once access-recommendation process formation by access user forms the model that Adds User
Train entry, specifically can use various ways, it is a kind of fairly simple it is preferable that, by these contents
With fixed form formation log recording, and sent at any time with log mode, the recipient of the daily record is according to day
The regulation of will form, is reduced to form or other data storage formats, and instruct according to user model
Parsed the need for white silk, extract one or several user models training entry, user model instruction
Practice the user characteristic data that records in entry, can as model training input data, user model instruction
In the personalized recommendation information for practicing bar program recording, access user does not make feedback, can be used as user
The negative sample of training, access user makes feedback, the positive sample that can be trained as user.
Step S104, training sample set is added by the model training entry that Adds User.
In this step, the model training entry that Adds User that previous step is formed is added into training sample set.Institute
Call training sample set, i.e., the set for the sample data trained for user model；User model is trained
Need to collect substantial amounts of user model training entry, could be realized to user model from substantial amounts of sample data
Significant adjustment, these training samples constantly accumulate, and are constantly used for user model and train, then finally
The user model that training is obtained will be more accurate.
In practical implementations, the model training entry that Adds User is the access process by accessing user each time
All access associated datas of middle formation, each is added up, and bar is trained by constantly accumulating user model
Mesh, in fact obtains the effect of big data.
The training sample set, is stored in the training sample set database of server side；The training sample
Collect the requirement according to model training, can typically use specific record format；The user model that will be increased newly
Train entry add training sample set process, specifically include the resolving to original entries, most at last its
In valid data with satisfactory form be stored in training sample set where database.
For example, the model training entry that Adds User of previous step formation uses logged, then by institute
State daily record and be sent to server side, server side is using predetermined journal format as foundation, to this daily record
The content of middle record is parsed, and the user model is trained to what the content in entry needed according to training data
User model training entry is formed, and is recorded the relevant position of training sample set database.
Step S105, user model training is carried out with current training sample set.
The current training sample set, refers to all effective instructions collected when starting user model training
Practice the data acquisition system of sample composition.
In the technical scheme that the application is provided, training data is constantly accumulated by each user's access process,
The content of its training sample set is the accumulative process of a dynamic.Therefore, when starting user model training every time,
Its training sample set can all change, i.e., sample data can increase；Mould is carried out using more sample datas
Type training, then the accuracy of model can also be lifted.Certainly, training sample set can also set what data were eliminated
Mechanism, for example, the sample data that can will build up on practice more than certain time is eliminated；Such as certain user's shopping
Correlated samples data and be that the year before, can not accurately reflect the situation of user, then these data
It can eliminate.
The user model training sample is trained for user model, can be used and be instructed newly-increased user model
White silk is adjusted to original user model, that is, carries out incremental training or the user model that will be increased newly
Training sample is added in original user model training sample, forms the user model training sample of a full dose,
And from the beginning user model is trained, i.e., full dose is trained.
Due to starting after user model training, training process needs the long period, and takes computing resource,
Therefore, under normal conditions, the model training sample that will not typically occur Adding User every time all starts a mould
Type training, but just start the model training in the case where meeting certain condition.I.e. before this step,
Increase by one judges whether the newly-increased user model training sample reaches the deterministic process of predetermined threshold value,
If so, then entering this step, otherwise, this step is not entered temporarily.
The certain condition, can include accumulated number condition and integration time condition, it may be considered that above-mentioned
Two conditions any one, two threshold conditions can also be considered simultaneously.
For example, in one constantly the electric business platform of generation transaction, it may be determined that reasonable time threshold value, example
Such as 24 hours, after the time threshold is reached, then start the training of new round user model, by this time
Newly-increased user model training sample is used in the training of this wheel.
For another example, in the fewer auction platform of a transaction, it may be determined that appropriate amount threshold, when new
When the entry accumulated number of the user model training sample of increasing exceedes the threshold value, then start user's mould of a new round
Type training, and newly-increased user model training sample is used in the training of this wheel.
Similar, for some platform, integration time threshold value and accumulated number threshold value, two can be set simultaneously
Person reaches condition, then starts user model training；Or, both any one reach condition, then start
User model is trained.
The user model training, can be realized using various ways, mainly pass through machine under the prior art
Device learning method realizes that user model is trained, and the machine learning method specifically used for example can be logistic regression
Method or gradient lifting traditional decision-tree.Letter is done to user model training process in logistic regression method below
Illustrate.
The logistic regression method includes：Training data is collected, feature extraction, Feature Selection, model training.
The training data collection is the user model training sample set of abovementioned steps formation.
The feature extraction, i.e., the data concentrated according to above-mentioned training sample are collected related to fit object
Various data.
The Feature Selection, weighs feature with correlation balancing method and (user characteristics number is come from the present embodiment
According to) degree of relevancy between fit object, and filter the feature that correlation is less than given threshold value；For example,
The present embodiment is used for clothes electric business, and given data that can be in training sample weighs the year of user
Age, sex character are with recommending the correlation between apparel brand, if correlation reaches default threshold requirement,
Then by the age, sex character is used as the feature related to apparel brand.
The model training stage, based on foregoing user model training data, to the number of users with correlation
According to fit regression model so that the difference of the desired value in the predicted value and training data that are obtained according to regression model
Away from minimum.I.e.：By constantly adjusting relevant parameter, make the predicted value that fitting is obtained (i.e. in the present embodiment
Personalized recommendation information) (i.e. user makes the personalization of feedback with the desired value in user model training data
Recommendation information) it is as consistent as possible.
In addition to above-mentioned logistic regression method, traditional decision-tree can also be lifted using gradient.
Above-mentioned user model training method uses algorithm ripe under prior art, is not belonging to the only of the present invention
Invasive part, is not described in detail herein.
Step S106, after the completion of the user model training, using the user model of the renewal obtained as working as
Preceding user model.
After previous step completes user model training, the user model obtained compares the user originally used
Model, the user model exactly updated；The user model of the renewal can substitute original user model at once
Used as active user's model.When performing the step S102, active user's model is exactly to pass through
This user model updated.
Using this method, data can be accumulated while carrying out model modification, and after model modification at once
Come into operation, new user model training data then accumulated under new user model, and go round and begin again,
Realize that positive improve each other of data accumulation and user model is circulated, it is rapid to improve user model quality,
Lift the accuracy of personalized recommendation information.
In above-mentioned first embodiment, relate to access the client at the access interface of user, with being used
The server of family model training；Pass through net connection between the two.The client provides the body for accessing user
Number is according to waiting to the server, and the user characteristic data that the server is stored according to these data queries is simultaneously
Client confession is sent to according to user characteristic data formation personalized recommendation information, and by personalized recommendation information
Displaying；The client, which is collected, accesses the user characteristic data obtained in user's access-recommendation process, personalization
The information such as recommendation information and feedback operation, forms the model training entry that Adds User, and be sent to service
Device, server trains entry according to the user model constantly accumulated, and machine carries out user model training in due course,
The active user's model updated is generated, and begins to use active user's model.
The user model of the step S103 trains the forming process of entry, can be completed, also may be used by client
To be to be collected to access the data that the access process of user is produced every time in real time by server on backstage, these data
In, personalized recommendation information is collected while sending, and user characteristic data can be sent in client
When accessing the identity of user, inquiry customer data base is obtained, and accesses user to personalized recommendation information
Feedback, the correlation circumstance provided from client；Finally, certain is once accessed-recommended in server-side
Above-mentioned all data acquisition systems of journey formation are a Data Entry, and are sent to the data of storage training sample set
Library storage.
The application second embodiment provides a kind of device for realizing online personalized recommendation；It refer to Fig. 2.
The device for the on-line system personalized recommendation that the present embodiment is provided, including：User characteristic data is extracted single
Member 201, personalized recommendation information provider unit 202, user model training entry formation unit 203, training
Sample set collects unit 204, user model training unit 205, user model updating block 206.
The user characteristic data extraction unit 201, the access request of user is accessed for receiving, and extracts institute
State the user characteristic data for accessing user.The user characteristic data at least includes：User Identity, and wrap
Include at least one in following any user characteristic data：Sex, age, transaction record.
The personalized recommendation information provider unit 202, for the user characteristics number according to the access user
According to, and there is provided personalized recommendation information for active user's model.The personalized recommendation information includes following
At least one of personalized recommendation entity：Commodity, shop, brand.
The user model training entry formation unit 203, the user for collecting the access user in real time is special
Levy data, make feedback for the personalized recommendation information that provides of access user, the access user
The personalized recommendation information of operation and the feedback operation done, and form the model training bar that Adds User
Mesh.The access user, which makes feedback operation, includes one of following operation：Click on, access, thumb up, collection,
It is predetermined, purchase.The one or two kinds of of data below is may also include in the training data entry：Obtaining should
The time point of record, resource-niche information.The user model training entry is preferred to use log mode record.
The training sample set collects unit 204, for the model training entry that Adds User to be added into training
Sample set.
The user model training unit 205, for carrying out user model training with current training sample set.Institute
State user model training and use machine learning method；Specific machine learning method can for example be returned using logic
Method or gradient is returned to lift traditional decision-tree.
The user model updating block 206, for after the completion of user model training, by what is obtained
The user model of renewal is used as active user's model.
In preferred embodiments, the device of the on-line system personalized recommendation is also alone including threshold decision,
For judging whether the newly-increased user model training sample reaches predetermined threshold value；If so, then starting institute
State user model training unit 205 and carry out user model training.
The application 3rd embodiment provides a kind of electronic equipment, and the electronic equipment includes：
Display；
Processor；
Memory, the program for storing the method for realizing on-line system personalized recommendation, equipment is powered simultaneously
After the program for the method for running the on-line system personalized recommendation, following step is performed：
The access request for accessing user is received, and extracts the user characteristic data of the access user；
According to the user characteristic data of the access user, and there is provided personalized recommendation for active user's model
Information；
The user characteristic data for collecting the access user in real time, the individual character provided for the access user
Change recommendation information, the access user make the personalized recommendation information of feedback operation and done it is anti-
Operation is presented, and the above of once access-recommendation process formation to access user is combined as one and is increased newly
User model training entry；
The newly-increased user model training entry is added into training sample set；
User model training is carried out with current training sample set；
After the completion of the user model training, the user model of the renewal obtained is regard as active user's model.
The application fourth embodiment provides a kind of online personalized recommendation for realizing above-mentioned first embodiment method
System；It refer to Fig. 3.
The online personalized system, including online subsystem 401, offline subsystem 402.
The online subsystem 401, for receiving the access request of user, and extracts user characteristic data, and
According to the user characteristic data and active user's model extracted, provide individual to the user for proposing access request
Property recommended entity；And, the user characteristic data of the access user is collected in real time, used for described access
The personalized recommendation information that family is provided, the access user are to the feedback letter of the personalized recommendation information
Breath, forms newly-increased user model training entry and sends；And, receive what the offline subsystem was provided
The user model of renewal.
The online subsystem realizes the function of real-time online, including showing interface and the real-time number of client
According to collection etc., the system emphasizes the real-time collection to all training datas, disposable complete collection related data
Form user model training entry, can thus avoid offline gather data need to access multiple databases with
And the drawbacks of partial data can not be collected in time.Online subsystem can include passing through in addition to client
The connected server of network.
In preferred scheme, the user model of the online formation of subsystem 401 trains entry with daily record shape
Formula recording and sending.
The offline subsystem 402, receives the newly-increased user model training that the online subsystem 401 is sent
Entry, and the newly-increased user model training entry is added into current training sample set；Using described current
Training sample set carries out user model training, regard the user model for training the renewal for completing to obtain as current use
Family model is sent out.
The offline subsystem by the newly-increased user model training entry add current training sample set it
Afterwards, judge whether the newly-increased user model training sample reaches predetermined threshold value at any time；If so, then opening
Begin the process that user model training is carried out using the current training sample set.
The offline subsystem 402, which is not handled, to work online, and can start user model training as needed at any time,
The work of on-line system is not influenceed.
Although the application is disclosed as above with preferred embodiment, it is not for limiting the application, Ren Heben
Art personnel are not being departed from spirit and scope, can make possible variation and modification,
Therefore the scope that the protection domain of the application should be defined by the application claim is defined.
In a typical configuration, computing device includes one or more processors (CPU), input/output
Interface, network interface and internal memory.
Internal memory potentially includes the volatile memory in computer-readable medium, random access memory
And/or the form, such as read-only storage (ROM) or flash memory (flash RAM) such as Nonvolatile memory (RAM).
Internal memory is the example of computer-readable medium.
1st, computer-readable medium include permanent and non-permanent, removable and non-removable media can be by
Any method or technique come realize information store.Information can be computer-readable instruction, data structure, journey
The module of sequence or other data.The example of the storage medium of computer includes, but are not limited to phase transition internal memory
(PRAM), static RAM (SRAM), dynamic random access memory (DRAM), other
The random access memory (RAM) of type, read-only storage (ROM), the read-only storage of electrically erasable
Device (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc read-only storage (CD-ROM),
Digital versatile disc (DVD) or other optical storages, magnetic cassette tape, tape magnetic rigid disk are stored or other
Magnetic storage apparatus or any other non-transmission medium, the information that can be accessed by a computing device available for storage.
Defined according to herein, computer-readable medium does not include non-temporary computer readable media (transitory
Media), such as the data-signal and carrier wave of modulation.
2nd, it will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer
Program product.Therefore, the application can use complete hardware embodiment, complete software embodiment or combine software
With the form of the embodiment of hardware aspect.Moreover, the application can be used wherein includes meter one or more
Calculation machine usable program code computer-usable storage medium (include but is not limited to magnetic disk storage,
CD-ROM, optical memory etc.) on the form of computer program product implemented.
Claims (12)
1. a kind of method of on-line system personalized recommendation, it is characterised in that including：
The access request for accessing user is received, and extracts the user characteristic data of the access user；
According to the user characteristic data of the access user, and there is provided personalized recommendation for active user's model
Information；
The user characteristic data for collecting the access user in real time, the individual character provided for the access user
Change recommendation information, the access user make the personalized recommendation information of feedback operation and done it is anti-
Feedback operation, and form the model training entry that Adds User；
The model training entry that Adds User is added into training sample set；
User model training is carried out with current training sample set；
After the completion of the user model training, the user model of the renewal obtained is regard as active user's model.
2. the method for on-line system personalized recommendation according to claim 1, it is characterised in that：
The user characteristic data at least includes：User Identity, and including following any user characteristics number
At least one in：Sex, age, transaction record.
3. the method for on-line system personalized recommendation according to claim 1, it is characterised in that：
The personalized recommendation information includes at least one of following personalized recommendation entity：Commodity, shop,
Brand.
4. the method for on-line system personalized recommendation according to claim 1, it is characterised in that：
The access user, which makes feedback operation, includes one of following operation：Click on, access, thumb up, collection,
It is predetermined, purchase.
5. the method for on-line system personalized recommendation according to claim 1, it is characterised in that：
Also include the one or two kinds of of data below in the training data entry：Obtain the time of the record
Point, resource-niche information.
6. the method for on-line system personalized recommendation according to claim 1, it is characterised in that by product
The tired model training entry that Adds User is added after training sample set, described with current training sample set
Before the step of carrying out user model training, following step is performed：
Whether the model training sample that Added User described in judging reaches predetermined threshold value；If so, then entering next
Step.
7. the method for on-line system personalized recommendation according to claim 1, it is characterised in that：It is described
User model training uses machine learning method.
8. the method for on-line system personalized recommendation according to claim 7, it is characterised in that described
Machine learning method is using logistic regression method or gradient lifting traditional decision-tree.
9. the method for on-line system personalized recommendation according to claim 1, it is characterised in that described
The model training entry that Adds User is recorded using log mode.
10. a kind of device of on-line system personalized recommendation, it is characterised in that including：
User characteristic data extraction unit, the access request of user is accessed for receiving, and extracts the access
The user characteristic data of user；
Personalized recommendation information provider unit, for the user characteristic data according to the access user, and
There is provided personalized recommendation information for active user's model；
User model training entry formation unit, for collect in real time it is described access user user characteristic data,
The institute of feedback operation is made for the personalized recommendation information of the access user offer, the access user
Personalized recommendation information and the feedback operation done are stated, and forms the model training entry that Adds User；
Training sample set collects unit, for the model training entry that Adds User to be added into training sample set；
User model training unit, for carrying out user model training with current training sample set；
User model updating block, for after the completion of user model training, by the renewal obtained
User model is used as active user's model.
11. a kind of electronic equipment, it is characterised in that the electronic equipment includes：
Display；
Processor；
Memory, the program for storing the method for realizing on-line system personalized recommendation, equipment is powered simultaneously
After the program for the method for running the on-line system personalized recommendation, following step is performed：
The access request for accessing user is received, and extracts the user characteristic data of the access user；
According to the user characteristic data of the access user, and there is provided personalized recommendation for active user's model
Information；
The user characteristic data for collecting the access user in real time, the individual character provided for the access user
Change recommendation information, the access user make the personalized recommendation information of feedback operation and done it is anti-
Operation is presented, and the above of once access-recommendation process formation to access user is combined as one and is increased newly
User model training entry；
The newly-increased user model training entry is added into training sample set；
User model training is carried out with current training sample set；
After the completion of the user model training, the user model of the renewal obtained is regard as active user's model.
12. a kind of system of online personalized recommendation, it is characterised in that including：It is online subsystem, offline
Subsystem；
The online subsystem, for receiving the access request of user, and extracts user characteristic data, and root
According to the user characteristic data and active user's model extracted, individual character is provided to the user for proposing access request
Change recommended entity；And, the user characteristic data of the access user is collected in real time, be the access user
The personalized recommendation information that there is provided, the access user to the feedback information of the personalized recommendation information,
Formation, which Adds User, model training entry and to be sent；And, receive the renewal that the offline subsystem is provided
User model；
The offline subsystem, receives the model training entry that Adds User that the online subsystem is sent, and
The model training entry that Adds User is added into current training sample set；Using the current training sample set
User model training is carried out, the user model for training the renewal for completing to obtain is sent out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610149438.5A CN107203518A (en) | 2016-03-16 | 2016-03-16 | Method, system and device, the electronic equipment of on-line system personalized recommendation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610149438.5A CN107203518A (en) | 2016-03-16 | 2016-03-16 | Method, system and device, the electronic equipment of on-line system personalized recommendation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107203518A true CN107203518A (en) | 2017-09-26 |
Family
ID=59904165
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610149438.5A Pending CN107203518A (en) | 2016-03-16 | 2016-03-16 | Method, system and device, the electronic equipment of on-line system personalized recommendation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107203518A (en) |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107705183A (en) * | 2017-09-30 | 2018-02-16 | 深圳乐信软件技术有限公司 | Recommendation method, apparatus, storage medium and the server of a kind of commodity |
| CN108009877A (en) * | 2017-11-24 | 2018-05-08 | 阿里巴巴集团控股有限公司 | Information mining method and device |
| CN108763342A (en) * | 2018-05-14 | 2018-11-06 | 北京比特智学科技有限公司 | Education resource distribution method and device |
| CN108803868A (en) * | 2018-04-13 | 2018-11-13 | 北京诺亦腾科技有限公司 | A kind of VR scenes experience cabin collecting method and device |
| CN108846687A (en) * | 2018-04-02 | 2018-11-20 | 平安科技（深圳）有限公司 | Client segmentation method, apparatus and storage medium |
| CN109146606A (en) * | 2018-07-09 | 2019-01-04 | 广州品唯软件有限公司 | A kind of brand recommended method, electronic equipment, storage medium and system |
| CN109299368A (en) * | 2018-09-29 | 2019-02-01 | 北京思路创新科技有限公司 | A kind of method and system for the intelligent personalized recommendation of environmental information resource AI |
| CN109325170A (en) * | 2018-08-06 | 2019-02-12 | 江西清华泰豪三波电机有限公司 | Material method for pushing and device |
| CN109413175A (en) * | 2018-10-19 | 2019-03-01 | 北京奇艺世纪科技有限公司 | A kind of information processing method, device and electronic equipment |
| WO2019061989A1 (en) * | 2017-09-30 | 2019-04-04 | 平安科技（深圳）有限公司 | Loan risk control method, electronic device and readable storage medium |
| CN110059170A (en) * | 2019-03-21 | 2019-07-26 | 北京邮电大学 | More wheels based on user's interaction talk with on-line training method and system |
| CN110059254A (en) * | 2019-04-24 | 2019-07-26 | 秒针信息技术有限公司 | A kind of message push method and device |
| CN110135979A (en) * | 2019-05-14 | 2019-08-16 | 极智(上海)企业管理咨询有限公司 | The matched method, apparatus of individual demand, medium and electronic equipment |
| CN110163693A (en) * | 2018-01-30 | 2019-08-23 | 哈尔滨学院 | A kind of electric business Method of Commodity Recommendation based on big data |
| WO2019205950A1 (en) * | 2018-04-28 | 2019-10-31 | K11集团有限公司 | User data collection system and information pushing method |
| CN110413888A (en) * | 2019-07-24 | 2019-11-05 | 腾讯科技（深圳）有限公司 | A kind of books recommended method and device |
| CN110428298A (en) * | 2019-07-15 | 2019-11-08 | 阿里巴巴集团控股有限公司 | A kind of shop recommended method, device and equipment |
| CN110472995A (en) * | 2019-07-08 | 2019-11-19 | 汉海信息技术(上海)有限公司 | To shop prediction technique, device, readable storage medium storing program for executing and electronic equipment |
| CN110674408A (en) * | 2019-09-30 | 2020-01-10 | 北京三快在线科技有限公司 | Service platform, and real-time generation method and device of training sample |
| CN111246056A (en) * | 2018-11-29 | 2020-06-05 | 佳能株式会社 | Information processing apparatus, control method therefor, image capturing apparatus, and storage medium |
| CN111461283A (en) * | 2020-03-18 | 2020-07-28 | 上海携程商务有限公司 | Automatic iteration operation and maintenance method, system, equipment and storage medium of AI model |
| CN111861605A (en) * | 2019-04-28 | 2020-10-30 | 阿里巴巴集团控股有限公司 | Business object recommendation method |
| CN112182359A (en) * | 2019-07-05 | 2021-01-05 | 腾讯科技（深圳）有限公司 | Feature management method and system of recommendation model |
| CN112598487A (en) * | 2021-02-20 | 2021-04-02 | 汇正(广州)企业管理咨询有限公司 | Enterprise informatization management method and system based on artificial intelligence |
| CN113378060A (en) * | 2021-07-01 | 2021-09-10 | 第四范式（北京）技术有限公司 | Resource recommendation method, device, equipment and medium |
| CN113573106A (en) * | 2020-04-28 | 2021-10-29 | 北京达佳互联信息技术有限公司 | Model updating method and device for multimedia information and server |
| CN115470397A (en) * | 2021-06-10 | 2022-12-13 | 腾讯科技（深圳）有限公司 | Content recommendation method and device, computer equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206751A (en) * | 2007-12-25 | 2008-06-25 | 北京科文书业信息技术有限公司 | Customer recommendation system based on data digging and method thereof |
| CN102800006A (en) * | 2012-07-23 | 2012-11-28 | 姚明东 | Real-time goods recommendation method based on customer shopping intention exploration |
| CN103020128A (en) * | 2012-11-19 | 2013-04-03 | 北京奇虎科技有限公司 | Method and device for data interaction with terminal device |
| CN103164434A (en) * | 2011-12-13 | 2013-06-19 | 阿里巴巴集团控股有限公司 | Method, device and system for obtaining real-time data |
| CN103678647A (en) * | 2013-12-20 | 2014-03-26 | Tcl集团股份有限公司 | Method and system for recommending information |
| CN104376058A (en) * | 2014-11-07 | 2015-02-25 | 华为技术有限公司 | User interest model updating method and device |
| CN104851025A (en) * | 2015-05-09 | 2015-08-19 | 湘南学院 | Case-reasoning-based personalized recommendation method for E-commerce website commodity |
-
2016
- 2016-03-16 CN CN201610149438.5A patent/CN107203518A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101206751A (en) * | 2007-12-25 | 2008-06-25 | 北京科文书业信息技术有限公司 | Customer recommendation system based on data digging and method thereof |
| CN103164434A (en) * | 2011-12-13 | 2013-06-19 | 阿里巴巴集团控股有限公司 | Method, device and system for obtaining real-time data |
| CN102800006A (en) * | 2012-07-23 | 2012-11-28 | 姚明东 | Real-time goods recommendation method based on customer shopping intention exploration |
| CN103020128A (en) * | 2012-11-19 | 2013-04-03 | 北京奇虎科技有限公司 | Method and device for data interaction with terminal device |
| CN103678647A (en) * | 2013-12-20 | 2014-03-26 | Tcl集团股份有限公司 | Method and system for recommending information |
| CN104376058A (en) * | 2014-11-07 | 2015-02-25 | 华为技术有限公司 | User interest model updating method and device |
| CN104851025A (en) * | 2015-05-09 | 2015-08-19 | 湘南学院 | Case-reasoning-based personalized recommendation method for E-commerce website commodity |
Non-Patent Citations (1)
| Title |
|---|
| 张轩荣: "让客户的行为留下痕迹", 《大数据时代，营销团队就该这样管》 * |
Cited By (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019061989A1 (en) * | 2017-09-30 | 2019-04-04 | 平安科技（深圳）有限公司 | Loan risk control method, electronic device and readable storage medium |
| CN107705183B (en) * | 2017-09-30 | 2021-04-27 | 深圳乐信软件技术有限公司 | Commodity recommendation method and device, storage medium and server |
| CN107705183A (en) * | 2017-09-30 | 2018-02-16 | 深圳乐信软件技术有限公司 | Recommendation method, apparatus, storage medium and the server of a kind of commodity |
| CN108009877A (en) * | 2017-11-24 | 2018-05-08 | 阿里巴巴集团控股有限公司 | Information mining method and device |
| CN110163693A (en) * | 2018-01-30 | 2019-08-23 | 哈尔滨学院 | A kind of electric business Method of Commodity Recommendation based on big data |
| CN108846687A (en) * | 2018-04-02 | 2018-11-20 | 平安科技（深圳）有限公司 | Client segmentation method, apparatus and storage medium |
| CN108803868A (en) * | 2018-04-13 | 2018-11-13 | 北京诺亦腾科技有限公司 | A kind of VR scenes experience cabin collecting method and device |
| WO2019205950A1 (en) * | 2018-04-28 | 2019-10-31 | K11集团有限公司 | User data collection system and information pushing method |
| CN110415065A (en) * | 2018-04-28 | 2019-11-05 | K11集团有限公司 | User data collection system and information-pushing method |
| CN108763342A (en) * | 2018-05-14 | 2018-11-06 | 北京比特智学科技有限公司 | Education resource distribution method and device |
| CN109146606B (en) * | 2018-07-09 | 2022-02-22 | 广州品唯软件有限公司 | Brand recommendation method, electronic equipment, storage medium and system |
| CN109146606A (en) * | 2018-07-09 | 2019-01-04 | 广州品唯软件有限公司 | A kind of brand recommended method, electronic equipment, storage medium and system |
| CN109325170A (en) * | 2018-08-06 | 2019-02-12 | 江西清华泰豪三波电机有限公司 | Material method for pushing and device |
| CN109299368A (en) * | 2018-09-29 | 2019-02-01 | 北京思路创新科技有限公司 | A kind of method and system for the intelligent personalized recommendation of environmental information resource AI |
| CN109413175B (en) * | 2018-10-19 | 2021-07-20 | 北京奇艺世纪科技有限公司 | Information processing method and device and electronic equipment |
| CN109413175A (en) * | 2018-10-19 | 2019-03-01 | 北京奇艺世纪科技有限公司 | A kind of information processing method, device and electronic equipment |
| CN111246056A (en) * | 2018-11-29 | 2020-06-05 | 佳能株式会社 | Information processing apparatus, control method therefor, image capturing apparatus, and storage medium |
| CN110059170B (en) * | 2019-03-21 | 2022-04-26 | 北京邮电大学 | Multi-turn dialogue online training method and system based on user interaction |
| CN110059170A (en) * | 2019-03-21 | 2019-07-26 | 北京邮电大学 | More wheels based on user's interaction talk with on-line training method and system |
| CN110059254A (en) * | 2019-04-24 | 2019-07-26 | 秒针信息技术有限公司 | A kind of message push method and device |
| CN111861605A (en) * | 2019-04-28 | 2020-10-30 | 阿里巴巴集团控股有限公司 | Business object recommendation method |
| CN110135979A (en) * | 2019-05-14 | 2019-08-16 | 极智(上海)企业管理咨询有限公司 | The matched method, apparatus of individual demand, medium and electronic equipment |
| CN112182359B (en) * | 2019-07-05 | 2024-03-15 | 深圳市雅阅科技有限公司 | Feature management method and system of recommendation model |
| CN112182359A (en) * | 2019-07-05 | 2021-01-05 | 腾讯科技（深圳）有限公司 | Feature management method and system of recommendation model |
| CN110472995A (en) * | 2019-07-08 | 2019-11-19 | 汉海信息技术(上海)有限公司 | To shop prediction technique, device, readable storage medium storing program for executing and electronic equipment |
| CN110428298A (en) * | 2019-07-15 | 2019-11-08 | 阿里巴巴集团控股有限公司 | A kind of shop recommended method, device and equipment |
| CN110413888A (en) * | 2019-07-24 | 2019-11-05 | 腾讯科技（深圳）有限公司 | A kind of books recommended method and device |
| CN110413888B (en) * | 2019-07-24 | 2024-05-10 | 腾讯科技（深圳）有限公司 | Book recommendation method and device |
| CN110674408A (en) * | 2019-09-30 | 2020-01-10 | 北京三快在线科技有限公司 | Service platform, and real-time generation method and device of training sample |
| CN110674408B (en) * | 2019-09-30 | 2021-06-04 | 北京三快在线科技有限公司 | Service platform, and real-time generation method and device of training sample |
| CN111461283A (en) * | 2020-03-18 | 2020-07-28 | 上海携程商务有限公司 | Automatic iteration operation and maintenance method, system, equipment and storage medium of AI model |
| CN113573106A (en) * | 2020-04-28 | 2021-10-29 | 北京达佳互联信息技术有限公司 | Model updating method and device for multimedia information and server |
| CN113573106B (en) * | 2020-04-28 | 2023-03-21 | 北京达佳互联信息技术有限公司 | Model updating method and device for multimedia information and server |
| CN112598487A (en) * | 2021-02-20 | 2021-04-02 | 汇正(广州)企业管理咨询有限公司 | Enterprise informatization management method and system based on artificial intelligence |
| CN115470397A (en) * | 2021-06-10 | 2022-12-13 | 腾讯科技（深圳）有限公司 | Content recommendation method and device, computer equipment and storage medium |
| CN115470397B (en) * | 2021-06-10 | 2024-04-05 | 腾讯科技（深圳）有限公司 | Content recommendation method, device, computer equipment and storage medium |
| CN113378060A (en) * | 2021-07-01 | 2021-09-10 | 第四范式（北京）技术有限公司 | Resource recommendation method, device, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107203518A (en) | Method, system and device, the electronic equipment of on-line system personalized recommendation | |
| US11107118B2 (en) | Management of the display of online ad content consistent with one or more performance objectives for a webpage and/or website | |
| JP5078910B2 (en) | Estimating advertising quality from observed user behavior | |
| CN103886487B (en) | Based on personalized recommendation method and the system of distributed B2B platform | |
| KR101315926B1 (en) | Using estimated ad qualities for ad filtering, ranking and promotion | |
| CN106372249B (en) | A kind of clicking rate predictor method, device and electronic equipment | |
| Kuo et al. | Building and evaluating a location-based service recommendation system with a preference adjustment mechanism | |
| US20190012392A1 (en) | Method and device for pushing information | |
| CN109934721A (en) | Finance product recommended method, device, equipment and storage medium | |
| CN103118111B (en) | Information push method based on data from a plurality of data interaction centers | |
| US20140032264A1 (en) | Data refining engine for high performance analysis system and method | |
| US20130238422A1 (en) | Automated Multivariate Behavioral Prediction | |
| US20130238425A1 (en) | Advertisement Selection Using Multivariate Behavioral Model | |
| AU2006332534A1 (en) | Predicting ad quality | |
| CN103049440A (en) | Recommendation processing method and processing system for related articles | |
| CN103678518A (en) | Method and device for adjusting recommendation lists | |
| CN110399563A (en) | A kind of preference pattern and customized information active recommended method based on user behavior analysis | |
| WO2018107102A1 (en) | Network interaction system | |
| CN106776860A (en) | One kind search abstraction generating method and device | |
| CA3062119A1 (en) | Method and device for setting sample weight, and electronic apparatus | |
| CN106469392A (en) | Select and recommend to show the method and device of object | |
| CN109918563A (en) | A method of the book recommendation based on public data | |
| CN111429214B (en) | Transaction data-based buyer and seller matching method and device | |
| US20190065611A1 (en) | Search method and apparatus | |
| JP6709775B2 (en) | Calculation device, calculation method, and calculation program |
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 |