CN109299370A - Multipair grade personalized recommendation method - Google Patents
Multipair grade personalized recommendation method Download PDFInfo
- Publication number
- CN109299370A CN109299370A CN201811172906.6A CN201811172906A CN109299370A CN 109299370 A CN109299370 A CN 109299370A CN 201811172906 A CN201811172906 A CN 201811172906A CN 109299370 A CN109299370 A CN 109299370A
- Authority
- CN
- China
- Prior art keywords
- user
- commodity
- collection
- preference
- multipair
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Abstract
The invention discloses a kind of multipair grade personalized recommendation methods, comprising: the implicit feedback information of user is extracted by internet platform;Determine that user to the preference of each commodity, so that the implicit feedback information of user is divided into positive feedback collection, negative-feedback collection and unknown collection, and regard these three set as training data according to the implicit feedback information of the user extracted;Optimize user to the preference of commodity using stochastic gradient descent algorithm and combined training data, thus preference of the user after being optimized to each commodity;It is ranked up from high to low according to the preference of commodity, several commodity in the top is done into the commercial product recommending liked as user to user.This method can excavate the potential interested merchandise news of user, and the commodity that user likes are recommended each user in the form of a list using the method for personalized recommendation.
Description
Technical field
The present invention relates to machine learning and recommender system field more particularly to a kind of multipair grade personalized recommendation methods.
Background technique
Collaborative filtering is one of most common algorithm in recommender system.What previous studies were more focused on is based on use
The collaborative filtering of family score data, but in life being difficult to obtain scoring of the user to commodity in more application scenarios.
For example, possessing the purchaser record of user on " day cat net ", the concern record of user is possessed on " microblogging " platform, in " love surprise
Possess the browsing record of user in skill ", the historical record of these users does not simultaneously include explicit score information, claims such
Data are the implicit feedback of user.Unlike the score data of user, implicit feedback contains only user for commodity
Positive and negative feedforward information largely cannot not be interpreted as user simply by the feedback information that commercial podium is observed and exist to these commodity
Negative-feedback, because user does not have found these commodity probably, rather than user does not like these commodity.
Since the recommender system based on implicit feedback lacks a large amount of negative-feedback information, especially the Sparse the case where
Under it is particularly evident, scholars expand a large amount of research work around this project, midpoint grade regression model and arrange grade
Sequence model achieves best recommendation effect.
Grade regression model is put using implicit feedback as user to the absolute preference value of commodity, and using a point grade Squared Error Loss letter
Number, which minimizes, carrys out approximated user to the absolute preference value of commodity.But it is very low to put grade regression model training effectiveness, uses in face of extensive
When the feedback data of family, preferable model can not be provided in effective time, and the experimental results showed that, point grade regression model for
The prediction effect of volatile data is poor, and recommendation effect is influenced bigger by initialization weight.
To grade order models using the preference relation of user and every a pair of of commodity as basic unit, to what is observed on platform
Commodity feedback information and the modeling of unobservable commodity feedback information, and the preference for attempting to maximize between commodity pair is assumed seemingly
Right function, to provide the interested items list of user.Bayes's personalized recommendation algorithm is most common using to grade row
The algorithm of sequence model, in recent years many research work are unfolded around Bayes's personalized recommendation algorithm and achieve good application
Value.But these to grade order models think user compared to the commodity of not no post-consumer prefer those given it is positive and negative
The commodity of feedback limit recommender system to have ignored the commodity that potential user likes in the commodity not being observed largely
Understanding and utilization to user preference.
Summary of the invention
The object of the present invention is to provide a kind of multipair grade personalized recommendation methods, can excavate the potential interested quotient of user
The commodity that user likes are recommended each user using the method for personalized recommendation by product information in the form of a list.
The purpose of the present invention is what is be achieved through the following technical solutions:
A kind of multipair grade personalized recommendation method, comprising:
The implicit feedback information of user is extracted by internet platform;
User is determined according to the implicit feedback information of the user extracted to the preference of each commodity, thus by user
Implicit feedback information be divided into positive feedback collection, negative-feedback collection and unknown collection, and regard these three set as training data;
Optimize user to the preference of commodity using stochastic gradient descent algorithm and combined training data, to obtain
The preference of user after optimization to each commodity;
It is ranked up from high to low according to the preference of commodity, several commodity in the top is done as user and are liked
Commercial product recommending to user.
As seen from the above technical solution provided by the invention, the quotient that each user does not generate buying behavior is treated with a certain discrimination
Product, not see preference of the user to this part commodity or do not like and be divided into two according to user, and go deep into having excavated use
Family loses the potential consumption demand of buying behavior because merchandise news is not known about, and finally provides and more reacts user preference degree
Commercial product recommending list.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, required use in being described below to embodiment
Attached drawing be briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this
For the those of ordinary skill in field, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of flow chart of multipair grade personalized recommendation method provided in an embodiment of the present invention;
Fig. 2 is that commodity data provided in an embodiment of the present invention divides schematic diagram;
Fig. 3 is the preference relation schematic diagram of multipair grade personalized ordering provided in an embodiment of the present invention.
Specific embodiment
With reference to the attached drawing in the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clear, complete
Ground description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based on this
The embodiment of invention, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, belongs to protection scope of the present invention.
The embodiment of the present invention provides a kind of multipair grade personalized recommendation method, as shown in Figure 1, it mainly includes walking as follows
It is rapid:
Step 1, the implicit feedback information that user is extracted by internet platform.
This step preferred embodiment is as follows:
Firstly, using crawler technology, user is crawled to the click record and purchaser record of commodity from online shopping platform, from
Social platform crawls user and records to the concern of other users, crawls user from online Media platform and remembers to the browsing of online Media
Record;Wherein, for social platform, regard other users as commodity, regard concern operation as purchase to commodity and operate;For
Line media platform regards online Media as commodity, regards browse operation of the user for online Media as purchase operation.
Then, data are pre-processed, constructs user-commodity matrix;Assuming that user is u, commodity i, user-commodity
Element (u, i) in matrix has recorded user u to the operation history of commodity i;Use ruiIndicate user u for the preference of commodity i
Degree, if user u has purchased commodity i, i.e. (u, i)=1, then it is assumed that user u expresses the positive feedback to commodity i, is denoted as rui
=1.
Step 2, determined according to the implicit feedback information of the user extracted user to the preference of each commodity, thus
The implicit feedback information of user is divided into positive feedback collection, negative-feedback collection and unknown collection, and these three are gathered as training number
According to.
In the embodiment of the present invention, all commodity set I are divided into positive feedback collection, negative-feedback collection and unknown collection.Specifically
For: 1) user u is bought into the set of commodity as positive feedback collectionNamely user is to the preference of these commodity
1.2) user had been observed that but simultaneously the set of non-purchased goods as negative-feedback collection3) positive feedback collection will be removed and born
Commodity set after feedback collection is as unknown collectionNamely do not know whether user u likes these commodity, because have very much can by user u
It can not see the information in relation to this part commodity.It should be noted negative-feedback collectionWith unknown collectionIt is all not have on platform
Observe the set of the commodity of feedback.
The preferred division mode of this step is as follows:
1) for online shopping platform, the commodity that user u is clicked and bought are included in positive feedback collectionUser u is clicked and is remembered
Non-purchased goods are included in negative-feedback collection in recordNamely user has been observed that this part commodity but does not select to buy them, because
And tend to no longer this part commercial product recommending to user;The user u commodity not clicked on finally are included in unknown collection
2) for social platform and online Media platform, the number that user is paid close attention to by other users and online Media quilt
The number of all users browsing selects sequence a part of user rearward and online Media as not respectively from how much being ranked up
Prevalence collection Ie;For user u, the other users of concern and the online Media of browsing are included in positive feedback collectionUnpopular collection IeWith
Positive feedback collectionDifference set(unpopular collection IePositive feedback collection) it is included in negative-feedback collectionUser u concern other users and
The online Media and difference set of browsingUnion, and all users and all online Medias, between difference set be then included in unknown collectionThat is, by " (all users and all online Medias)-[(difference set) ∪ (and user u concern other users and browsing
Line media)] " result be included in unknown collection
As shown in Fig. 2, dividing schematic diagram for commodity data.User John like film " The Dark Knight " and
" Alien ", thus positive feedback collection will be included inThe film of some few people's viewings represents unpopular collection Ie(shadow part
Point), then for user John, the unpopular collection I of dash areaeWith the difference set of " The Dark Knight " and " Alien "
Negative-feedback collection will be included in(i.e. unpopular collection IeThe middle data removed except Alien), it is believed that user John is to this part film
Lose interest in.Whether user John, which likes, to be judged for other films, thus is all included in unknown collectionIn.
Step 3, using stochastic gradient descent algorithm and combined training data optimize user to the preference of commodity, from
And the user after being optimized is to the preference of each commodity.
By positive feedback collectionMiddle user is denoted as 1 to the preference of each commodity, by negative-feedback collectionMiddle user is to each quotient
The preference of product is denoted as 0, by unknown collectionPreference question mark of the middle user to each commodity? to indicate.
As shown in figure 3, is sampled by 6 different commodity in commodity set I, is respectively labeled as i, j, p by user u,
P ', q, q ', wherein i, p, p ' belong to positive feedback collectionJ, q ' belong to negative-feedback collectionQ belongs to unknown collectionThen for user
U, to the preference of commodity are as follows: rui=rup=rup′=1, ruj≈ruq′≈ 0, ruq=?, 0≤? ≤ 1;R therein indicates preference
Degree degree, two subscripts are corresponding in turn to and user and commodity.
According to user u for the preference of commodity, the preference relation that can provide multipair grade personalized ordering algorithm is assumed: because
Positive feedback is expressed to commodity i for user u, and user u does not like commodity j and commodity q ' probably, so no matter
Commodity q is the interested commodity of user u, can all there is preference relation rui-ruj≥ruq-ruq′.Similarly, user u likes simultaneously
Commodity p and p ', thus user should be for the difference of their hobby it is smaller, that is, there is following preference relation ruq-ruq′
≥rup-rup′.Utilize mark ruij=rui-rujUser u is indicated to the difference of preference between commodity i and commodity j, according to more
Preference to grade personalized recommendation algorithm (MPR) is it is assumed that r can be summarized asuij≥ruqq′≥rupp′, then for all users
U has following likelihood function:
Above-mentioned likelihood function contains 3 different commodity pair, deeply understands each user for purchase and does not purchase
Buy the preference relation between commodity.All in all, the preference difference between two commodity bought by user u is not less than two
Preference difference between the commodity bought by user u, the latter again less than one by the user u commodity bought and one not by
Preference difference between the commodity that user u was bought.By user to the ratio of difference preference's relationship difference between multiple commodity pair
Compared with having excavated user may interested commodity to non-purchased goods concentration.
By ruij≥ruqq′,ruqq′≥rupp′It indicates are as follows:
λ(ruij-ruqq′)+(1-λ)(ruqq′-rupp′);
Wherein, λ is the balance factor that target is assumed for balancing two preferences, and above formula is abbreviated asAnd it utilizes
Following formula carrys out approximate evaluation probability value Pr ():
Then for user u, the preference hypothesis of multipair grade personalized recommendation algorithm is write a Chinese character in simplified form are as follows:
Based on the above rule, the likelihood function for optimizing multipair grade personalized recommendation algorithm is indicated are as follows:
Wherein, Θ={ Uu·∈R1×d,Vi·∈R1×d,bi∈ R, u ∈ U, i ∈ I } it is the parameter that model needs to learn, Uu·It is
The feature vector of user u, V are describedi·It is the feature vector for describing commodity i, biIt is the offset of commodity i feature vector, commodity set I
The offset of middle product features vector is denoted as R, and d is the dimension of feature vector;R (Θ) is in order to avoid over-fitting in the training process
And the regularization term being arranged, R (Θ)=∑u∈U∑t∈S[αu‖Uu·‖2+αv‖Vt·‖2+βv‖bt·‖2], S={ i, j, p, p ', q, q ' }
It is the sampled sample of each round training;Ln MPR is the log-likelihood function of multipair grade personalized recommendation algorithm, is indicated are as follows:
Above-mentioned likelihood function is optimized using stochastic gradient descent algorithm (SGD), each round iterative process selects a record,
It contains user a u, 6 different commodity i, j, p, p ', q, q ' and optimization is reached according to the parameter of gradient information more new model
Purpose, final majorized function indicates are as follows:
After obtaining gradient signal, model parameter is updated by following formula:
In above formula, γ > 0 indicates learning rate, the updated model parameter of Θ ' expression, the parameter type and Θ for being included
It is identical, a slash is increased here for the model parameter distinguished before and after updating, thus for Θ, in actual operation, epicycle updates
Model parameter Θ ' afterwards makees the training of next round by Θ is assigned to.
Step 4 is ranked up from high to low according to the preference of commodity, using several commodity in the top as user
The commercial product recommending liked is done to user.
By learning process before, fancy grade of the available user to all commodity, the quotient comprising shopping platform
Product, the other users of social platform, the online Media of online Media platform.Degree sequence knot is liked to commodity according to user
Fruit, the favorite several commercial product recommendings of selection target user are to target user;For all users all in accordance with above-mentioned steps 1~
The mode of step 4 is handled, to realize the purpose of personalized recommendation to each user.
Above scheme of the embodiment of the present invention divides commodity set according to the history mutual information of user and commodity, passes through ratio
Compared with the relationship of preference difference between user and commodity pair, user has deeply been excavated for not buying the potential shopping need of commodity,
The preference for finally optimizing multipair grade personalized recommendation algorithm based on stochastic gradient descent algorithm more meets use it is assumed that recommending user
The items list of family demand.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment can
The mode of necessary general hardware platform can also be added to realize by software by software realization.Based on this understanding,
The technical solution of above-described embodiment can be embodied in the form of software products, which can store non-easy at one
In the property lost storage medium (can be CD-ROM, USB flash disk, mobile hard disk etc.), including some instructions are with so that a computer is set
Standby (can be personal computer, server or the network equipment etc.) executes method described in each embodiment of the present invention.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Within the technical scope of the present disclosure, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with the protection model of claims
Subject to enclosing.
Claims (4)
1. a kind of multipair grade personalized recommendation method characterized by comprising
The implicit feedback information of user is extracted by internet platform;
User is determined according to the implicit feedback information of the user extracted to the preference of each commodity, thus by the hidden of user
Formula feedback information is divided into positive feedback collection, negative-feedback collection and unknown collection, and regard these three set as training data;
Optimize user to the preference of commodity using stochastic gradient descent algorithm and combined training data, to be optimized
Preference of the user afterwards to each commodity;
It is ranked up from high to low according to the preference of commodity, several commodity in the top is done into the quotient liked as user
Product recommend user.
2. a kind of multipair grade personalized recommendation method according to claim 1, which is characterized in that described flat by internet
Platform extract user implicit feedback information include:
Firstly, crawling user to the click record and purchaser record of commodity, from social activity from online shopping platform using crawler technology
Platform crawls user and records to the concern of other users, crawls user from online Media platform and records to the browsing of online Media;
Wherein, for social platform, regard other users as commodity, regard concern operation as purchase to commodity and operate;For online
Media platform regards online Media as commodity, regards browse operation of the user for online Media as purchase operation;
Then, data are pre-processed, constructs user-commodity matrix;Assuming that user is u, commodity i, user-commodity matrix
In element (u, i) have recorded user u to the operation history of commodity i;Use ruiIndicate user u for commodity i preference,
If user u has purchased commodity i, i.e. (u, i)=1, then it is assumed that user u expresses the positive feedback to commodity i, is denoted as rui=1.
3. a kind of multipair grade personalized recommendation method according to claim 2, which is characterized in that the basis was extracted
The implicit feedback information of user determines user to the preference of each commodity, so that the implicit feedback information of user is divided into
Positive feedback collection, negative-feedback collection and unknown collection include:
User u the set of commodity has been bought into as positive feedback collectionNamely user is 1 to the preference of these commodity;It will
User have been observed that but simultaneously the set of non-purchased goods as negative-feedback collectionAfter positive feedback collection and negative-feedback collection being removed
Commodity set as unknown collection
For online shopping platform, the commodity that user u is clicked and bought are included in positive feedback collectionUser u is clicked in record not
The commodity of purchase are included in negative-feedback collectionThe user u commodity not clicked on are included in unknown collection
For social platform and online Media platform, the number that user is paid close attention to by other users and online Media are useful by institute
The number of family browsing selects a part of user and the unpopular collection of online Media conduct to sort rearward respectively from how much being ranked up
Ie;For user u, the other users of concern and the online Media of browsing are included in positive feedback collectionUnpopular collection IeWith positive feedback
CollectionDifference setIt is included in negative-feedback collectionThe other users of user u concern and the online Media and difference set of browsingUnion,
With all users and all online Medias, between difference set be then included in unknown collection
4. a kind of multipair grade personalized recommendation method according to claim 3, which is characterized in that described to use stochastic gradient
Descent algorithm and combined training data include: to optimize user to the preference of commodity
By positive feedback collectionMiddle user is denoted as 1 to the preference of each commodity, by negative-feedback collectionMiddle user is to each commodity
Preference is denoted as 0, by unknown collectionPreference question mark of the middle user to each commodity? to indicate;
For user u, 6 different commodity are sampled in commodity set I, are respectively labeled as i, j, p, p ', q, q ', wherein i,
P, p ' belong to positive feedback collectionJ, q ' belong to negative-feedback collectionQ belongs to unknown collectionThen for user u, to the preference journey of commodity
Degree are as follows: rui=rup=rup′=1, ruj≈ruq′≈ 0, ruq=?, 0≤? ≤ 1;R therein indicates preference degree, under two
Mark is corresponding in turn to and user and commodity;Utilize mark ruij=rui-rujIndicate user u to preference between commodity i and commodity j
Difference, according to the preference of multipair grade personalized recommendation algorithm MPR it is assumed that there is ruij≥ruqq′≥rupp′, then for all
User U has following likelihood function:
By ruij≥ruqq′, ruqq′≥rupp′It indicates are as follows:
λ(ruij-ruqq′)+(1-λ)(ruqq′-rupp′);
Wherein, λ is the balance factor that target is assumed for balancing two preferences, and above formula is abbreviated asAnd utilize following formula
Carry out approximate evaluation probability value Pr ():
Then for user u, the preference hypothesis of multipair grade personalized recommendation algorithm is write a Chinese character in simplified form are as follows:
It is indicated to optimize the likelihood function of multipair grade personalized recommendation algorithm are as follows:
Wherein, Θ={ Uu·∈R1×d, Vi·∈R1×d, bi∈ R, u ∈ U, i ∈ I } it is the parameter that model needs to learn, Uu·It is description
The feature vector of user u, Vi·It is the feature vector for describing commodity i, biIt is the offset of commodity i feature vector, quotient in commodity set I
The offset of product feature vector is denoted as R, and d is the dimension of feature vector;R (Θ) is regularization term;Ln MPR is that multipair grade is personalized
The log-likelihood function of proposed algorithm indicates are as follows:
Above-mentioned likelihood function is optimized using stochastic gradient descent algorithm, each round iterative process selects a record, contains one
A user u, 6 different commodity i, j, p, p ', q, q ' achieve the purpose that optimization according to the parameter of gradient information more new model,
Final majorized function indicates are as follows:
Wherein, S={ i, j, p, p ', q, q ' };
After obtaining gradient signal, model parameter is updated by following formula:
In above formula, γ > 0 indicates learning rate, the updated model parameter of Θ ' expression.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811172906.6A CN109299370B (en) | 2018-10-09 | 2018-10-09 | Multi-pair level personalized recommendation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811172906.6A CN109299370B (en) | 2018-10-09 | 2018-10-09 | Multi-pair level personalized recommendation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109299370A true CN109299370A (en) | 2019-02-01 |
CN109299370B CN109299370B (en) | 2022-03-01 |
Family
ID=65161832
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811172906.6A Active CN109299370B (en) | 2018-10-09 | 2018-10-09 | Multi-pair level personalized recommendation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109299370B (en) |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109934681A (en) * | 2019-03-13 | 2019-06-25 | 中国科学技术大学 | The recommended method of user's commodity interested |
CN110059251A (en) * | 2019-04-22 | 2019-07-26 | 郑州大学 | Collaborative filtering recommending method based on more relationship implicit feedback confidence levels |
CN110084670A (en) * | 2019-04-15 | 2019-08-02 | 东北大学 | A kind of commodity on shelf combined recommendation method based on LDA-MLP |
CN110348895A (en) * | 2019-06-29 | 2019-10-18 | 北京淇瑀信息科技有限公司 | A kind of personalized recommendation method based on user tag, device and electronic equipment |
CN110362753A (en) * | 2019-04-10 | 2019-10-22 | 深思考人工智能机器人科技(北京)有限公司 | A kind of personalized neural network recommendation method and system based on user concealed feedback |
CN110377783A (en) * | 2019-06-05 | 2019-10-25 | 深圳大学 | A kind of method, apparatus and computer equipment that audio-video is recommended |
CN110473543A (en) * | 2019-09-25 | 2019-11-19 | 北京蓦然认知科技有限公司 | A kind of audio recognition method, device |
CN110489645A (en) * | 2019-08-15 | 2019-11-22 | 深圳市云积分科技有限公司 | A kind of brand association consumer orients the information processing method and device of marketing |
CN110503531A (en) * | 2019-08-30 | 2019-11-26 | 中国科学技术大学 | The dynamic social activity scene recommended method of timing perception |
CN110544129A (en) * | 2019-09-05 | 2019-12-06 | 创新奇智(青岛)科技有限公司 | Personalized recommendation method for social e-commerce users |
CN110727872A (en) * | 2019-10-21 | 2020-01-24 | 深圳微品致远信息科技有限公司 | Method and device for mining ambiguous selection behavior based on implicit feedback |
CN111339327A (en) * | 2020-02-20 | 2020-06-26 | 北京达佳互联信息技术有限公司 | Work recommendation method and device, server and storage medium |
CN111523940A (en) * | 2020-04-23 | 2020-08-11 | 华中科技大学 | Deep reinforcement learning-based recommendation method and system with negative feedback |
CN111598644A (en) * | 2020-04-01 | 2020-08-28 | 华瑞新智科技(北京)有限公司 | Article recommendation method, device and medium |
CN114637922A (en) * | 2022-05-17 | 2022-06-17 | 浙江大学滨江研究院 | User preference recommendation method and device based on neural network set operation |
CN114780862A (en) * | 2022-06-21 | 2022-07-22 | 达而观数据(成都)有限公司 | User interest vector extraction method, extraction model and computer system |
CN114997956A (en) * | 2022-06-14 | 2022-09-02 | 杭州洋驼网络科技有限公司 | Mother and infant product intelligent recommendation system based on big data |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101383942A (en) * | 2008-08-01 | 2009-03-11 | 深圳市天威视讯股份有限公司 | Hidden customer characteristic extracting method and television program recommendation method and system |
CN102982107A (en) * | 2012-11-08 | 2013-03-20 | 北京航空航天大学 | Recommendation system optimization method with information of user and item and context attribute integrated |
CN103617540A (en) * | 2013-10-17 | 2014-03-05 | 浙江大学 | E-commerce recommendation method of tracking user interest changes |
US20160379257A1 (en) * | 2012-07-17 | 2016-12-29 | Yahoo! Inc. | Processing data across network elements |
CN107766439A (en) * | 2017-09-21 | 2018-03-06 | 汉鼎宇佑互联网股份有限公司 | A kind of personalized recommendation method of fusion structure feature and implicit feedback |
CN108460619A (en) * | 2018-01-16 | 2018-08-28 | 杭州电子科技大学 | A kind of fusion shows the Collaborative Recommendation model of implicit feedback |
-
2018
- 2018-10-09 CN CN201811172906.6A patent/CN109299370B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101383942A (en) * | 2008-08-01 | 2009-03-11 | 深圳市天威视讯股份有限公司 | Hidden customer characteristic extracting method and television program recommendation method and system |
US20160379257A1 (en) * | 2012-07-17 | 2016-12-29 | Yahoo! Inc. | Processing data across network elements |
CN102982107A (en) * | 2012-11-08 | 2013-03-20 | 北京航空航天大学 | Recommendation system optimization method with information of user and item and context attribute integrated |
CN103617540A (en) * | 2013-10-17 | 2014-03-05 | 浙江大学 | E-commerce recommendation method of tracking user interest changes |
CN107766439A (en) * | 2017-09-21 | 2018-03-06 | 汉鼎宇佑互联网股份有限公司 | A kind of personalized recommendation method of fusion structure feature and implicit feedback |
CN108460619A (en) * | 2018-01-16 | 2018-08-28 | 杭州电子科技大学 | A kind of fusion shows the Collaborative Recommendation model of implicit feedback |
Non-Patent Citations (1)
Title |
---|
印鉴等: "基于大规模隐式反馈的个性化推荐", 《软件学报》 * |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109934681A (en) * | 2019-03-13 | 2019-06-25 | 中国科学技术大学 | The recommended method of user's commodity interested |
CN109934681B (en) * | 2019-03-13 | 2022-01-11 | 中国科学技术大学 | Recommendation method for user interested goods |
CN110362753A (en) * | 2019-04-10 | 2019-10-22 | 深思考人工智能机器人科技(北京)有限公司 | A kind of personalized neural network recommendation method and system based on user concealed feedback |
CN110084670A (en) * | 2019-04-15 | 2019-08-02 | 东北大学 | A kind of commodity on shelf combined recommendation method based on LDA-MLP |
CN110084670B (en) * | 2019-04-15 | 2022-03-25 | 东北大学 | Shelf commodity combination recommendation method based on LDA-MLP |
CN110059251B (en) * | 2019-04-22 | 2022-10-28 | 郑州大学 | Collaborative filtering recommendation method based on multi-relation implicit feedback confidence |
CN110059251A (en) * | 2019-04-22 | 2019-07-26 | 郑州大学 | Collaborative filtering recommending method based on more relationship implicit feedback confidence levels |
CN110377783A (en) * | 2019-06-05 | 2019-10-25 | 深圳大学 | A kind of method, apparatus and computer equipment that audio-video is recommended |
CN110348895A (en) * | 2019-06-29 | 2019-10-18 | 北京淇瑀信息科技有限公司 | A kind of personalized recommendation method based on user tag, device and electronic equipment |
CN110489645A (en) * | 2019-08-15 | 2019-11-22 | 深圳市云积分科技有限公司 | A kind of brand association consumer orients the information processing method and device of marketing |
CN110503531A (en) * | 2019-08-30 | 2019-11-26 | 中国科学技术大学 | The dynamic social activity scene recommended method of timing perception |
CN110503531B (en) * | 2019-08-30 | 2022-04-19 | 中国科学技术大学 | Dynamic social scene recommendation method based on time sequence perception |
CN110544129A (en) * | 2019-09-05 | 2019-12-06 | 创新奇智(青岛)科技有限公司 | Personalized recommendation method for social e-commerce users |
CN110473543B (en) * | 2019-09-25 | 2020-11-27 | 北京蓦然认知科技有限公司 | Voice recognition method and device |
CN110473543A (en) * | 2019-09-25 | 2019-11-19 | 北京蓦然认知科技有限公司 | A kind of audio recognition method, device |
CN110727872A (en) * | 2019-10-21 | 2020-01-24 | 深圳微品致远信息科技有限公司 | Method and device for mining ambiguous selection behavior based on implicit feedback |
CN111339327A (en) * | 2020-02-20 | 2020-06-26 | 北京达佳互联信息技术有限公司 | Work recommendation method and device, server and storage medium |
CN111598644A (en) * | 2020-04-01 | 2020-08-28 | 华瑞新智科技(北京)有限公司 | Article recommendation method, device and medium |
CN111598644B (en) * | 2020-04-01 | 2023-05-02 | 华瑞新智科技(北京)有限公司 | Article recommendation method, device and medium |
CN111523940A (en) * | 2020-04-23 | 2020-08-11 | 华中科技大学 | Deep reinforcement learning-based recommendation method and system with negative feedback |
CN114637922B (en) * | 2022-05-17 | 2022-08-23 | 浙江大学滨江研究院 | User preference recommendation method and device based on neural network set operation |
CN114637922A (en) * | 2022-05-17 | 2022-06-17 | 浙江大学滨江研究院 | User preference recommendation method and device based on neural network set operation |
CN114997956A (en) * | 2022-06-14 | 2022-09-02 | 杭州洋驼网络科技有限公司 | Mother and infant product intelligent recommendation system based on big data |
CN114997956B (en) * | 2022-06-14 | 2023-01-10 | 杭州洋驼网络科技有限公司 | Mother and infant product intelligent recommendation system based on big data |
CN114780862B (en) * | 2022-06-21 | 2022-09-23 | 达而观数据(成都)有限公司 | User interest vector extraction method, extraction model and computer system |
CN114780862A (en) * | 2022-06-21 | 2022-07-22 | 达而观数据(成都)有限公司 | User interest vector extraction method, extraction model and computer system |
Also Published As
Publication number | Publication date |
---|---|
CN109299370B (en) | 2022-03-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109299370A (en) | Multipair grade personalized recommendation method | |
CN107330115B (en) | Information recommendation method and device | |
CN107563841B (en) | Recommendation system based on user score decomposition | |
CN111310063B (en) | Neural network-based article recommendation method for memory perception gated factorization machine | |
US11715132B2 (en) | Adaptive and recursive system and method | |
WO2021262530A1 (en) | Systems and methods for deep learning model based product matching using multi modal data | |
CN112313697A (en) | System and method for generating interpretable description-based recommendations describing angle augmentation | |
CN109190030B (en) | Implicit feedback recommendation method fusing node2vec and deep neural network | |
US11727014B2 (en) | Dynamic filter recommendations | |
CN105518665B (en) | Search result amplification | |
US11386301B2 (en) | Cluster and image-based feedback system | |
CN103198089A (en) | Information processing apparatus, information processing method, and program | |
US9767204B1 (en) | Category predictions identifying a search frequency | |
CN112380453B (en) | Article recommendation method and device, storage medium and equipment | |
CN113918833B (en) | Product recommendation method realized through graph convolution collaborative filtering of social network relationship | |
CN113918834B (en) | Graph convolution collaborative filtering recommendation method fusing social relations | |
US11651450B2 (en) | Simulated network system and method for relating users of real-world e-commerce and other user network systems to information | |
US20210118020A1 (en) | Price-Based User Feedback System | |
US10474670B1 (en) | Category predictions with browse node probabilities | |
CN103150667A (en) | Individualization recommending method based on mechanical structure | |
CN113918832A (en) | Graph convolution collaborative filtering recommendation system based on social relationship | |
CN112699310A (en) | Cold start cross-domain hybrid recommendation method and system based on deep neural network | |
CN109934681B (en) | Recommendation method for user interested goods | |
CN114579852A (en) | Recommendation method based on graph convolution network and attention mechanism | |
Wu et al. | Modeling uncertainty driven curiosity for social recommendation |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |