CN109886779A - A kind of intelligence commercial product recommending system - Google Patents
A kind of intelligence commercial product recommending system Download PDFInfo
- Publication number
- CN109886779A CN109886779A CN201910089640.7A CN201910089640A CN109886779A CN 109886779 A CN109886779 A CN 109886779A CN 201910089640 A CN201910089640 A CN 201910089640A CN 109886779 A CN109886779 A CN 109886779A
- Authority
- CN
- China
- Prior art keywords
- user
- matrix
- article
- preference
- commodity
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of intelligent commercial product recommending systems, including sequentially connected information initializing module, merchandise display module and adjustment module, the information initializing module is used to construct the comment document sets for trusting matrix, consumer articles between rating matrix, user and the user of user and article, preference matrix of the initialising subscriber to article;K before preference value comes commodity are calculated by preference matrix of the user to article for merchandise display module current scene for identification, and by merchandise news displaying in the software page, and collect user to the feedback information of commodity;The adjustment module is for adjusting rating matrix, trusting matrix, comment document sets and preference matrix.The present invention realizes the utilization of the fusion to multiple data sources, is one items list of user-customized recommended, and adjusted in real time according to feedback information to the items list that user likes, improves the recommendation quality of system.
Description
Technical field
The present invention relates to technical field of information processing more particularly to a kind of intelligent commercial product recommending systems.
Background technique
The function of having hobby to recommend in existing website, searches out use by a series of core algorithms of its internal system
The possible interested news in family or commodity, and it is shown in the forward position of homepage.It judges that the method for user preference substantially has two
Kind, one is using user to the scoring information of article, and the method based on collaborative filtering finds the neighbour with similar marking behavior
It occupies, and the commercial product recommending that the neighbours are liked is to user;Secondly for using user's registration information and commodity essential information, by interior
The recommended method of appearance recommends its interested article for user.Current each website is all more similar in recommended method, i.e., first
Commodity related data sources are first analyzed, and calculate user to the preference of each commodity.When needing to user's Recommendations, obtain
The user calculated gives to the preference matrix of article and recommends number K, by the highest preceding K commercial product recommending of preference value to use
Family.
Existing recommended method haves the shortcomings that as follows: it is insufficient to the excavation of preference between user and article, therefore
Finally recommend the commodity of user and commodity that user really likes there are larger gaps.Its scoring based on user and article is closed
System makes recommendation, has ignored other data sources, such as social information, score information, and is recommending its interested quotient for user
When product list, the ordering relation between commodity is had ignored.Recommend all to be to obtain analog result using identical data every time, ignore
The dynamic of user preference, when Recommendations and user preference are not met, system, which cannot voluntarily be adjusted, recommends that more its next time
It is accurate to add.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of intelligent commercial product recommending system, the system energy
User, the essential information of article and user are enough collected to the interbehavior data of article, and data are analyzed and processed, is dug
User preference is dug, system is improved and is the sequence quality of user's Recommendations list, and its interested personalization is generated for user
Commodity library voluntarily adjusts the items list of recommendation according to the preference of user, carries out personalized recommendation for user, improves and recommend matter
Amount.
In order to achieve the above objectives, the present invention adopts the following technical scheme:
A kind of intelligence commercial product recommending system, including sequentially connected information initializing module, merchandise display module and adjustment
Module, the information initializing module be used to construct trust matrix between rating matrix, user and the user of user and article,
The comment document sets of consumer articles, preference matrix of the initialising subscriber to article;The merchandise display module is current for identification
K before preference value comes commodity are calculated by preference matrix of the user to article, and merchandise news is shown for scene
The software page, and user is collected to the feedback information of commodity;The adjustment module is for adjusting rating matrix, trust matrix, commenting
By document sets and preference matrix.
The information initializing module includes acquiring unit and computing unit, and the acquiring unit is for obtaining commodity correlation
Data source;The computing unit be used to calculate trust strength TN [u, v] between user u and user v, similarity TS [u, v],
Comprehensive degree of belief T [u, v], and calculate and update user to the preference matrix RP of article.
The form for the relevant data source of commodity that the acquiring unit obtains are as follows:
P=Ui | i=1,2,3 ...
Q=Vj | j=1,2,3 ...
R=Rij | i ∈ U, j ∈ V }
S=Sij | i, j ∈ U }
RP=RPij | i ∈ U, j ∈ V }
Wherein: P is user's collection, and Ui is the user in user's set;Q is article collection, and Vj is the article in article set;R
For rating matrix, Rij indicates user i to the score value of article j;S is social matrix, and Sij is that user i closes the trust of user j
System;RP is preference matrix, and RPij indicates the score value that user i predicts article j.
The merchandise display module includes display unit and feedback unit;The display unit is used for from user to article
The commodity that K before preference value comes are obtained in preference matrix, are shown in the page;The feedback unit is newest for collecting
User analyzes these data to the interbehavior data of commodity, adjusts weight in real time.
Compared with prior art, the present invention has the advantage that:
Information initializing module of the present invention obtains the related data source information of user and commodity using its acquiring unit, and
Engine of its computing unit as system is calculated using the sort recommendations algorithm proposed by the present invention based on Multi-source Information Fusion
Preference matrix of the user to article.Merchandise display module then will be before user preference value comes in preference matrix using its display unit
K commodity are shown, and feedback behavior of the displayed page user to commodity is collected using the feedback unit in module.
Adjustment module then obtains latest data from the feedback unit in merchandise display module using its basic data adjustment unit to update
Data source, and the computing unit in initialization module is called to update user to the preference matrix of article to calculate.These three modules
An entirety is formed, is interacted, can accurately be excavated the preference of user, improve recommendation quality.
Detailed description of the invention
Fig. 1 is the composed structure schematic diagram of intelligent commercial product recommending system of the invention.
Specific embodiment
With reference to the accompanying drawing, technical scheme in the embodiment of the invention is clearly and completely described.
As shown in Figure 1, a kind of intelligence commercial product recommending system, including sequentially connected information initializing module 1, merchandise display
Module 2 and adjustment module 3, the information initializing module 1 be used for construct user and article rating matrix, user and user it
Between trust matrix, consumer articles comment document sets, preference matrix of the initialising subscriber to article;The merchandise display module
K before preference value comes commodity, and general are calculated by preference matrix of the user to article for 2 current scenes for identification
Merchandise news is shown in the software page;The adjustment module 3 for adjust rating matrix, trust matrix, comment document sets and partially
Good matrix.
The information initializing module 1 includes acquiring unit and computing unit, and the acquiring unit is for obtaining commodity phase
The data source of pass;The computing unit be used to calculate trust strength TN [u, v] between user u and user v, similarity TS [u,
V], comprehensive degree of belief T [u, v], and calculate and update user to the preference matrix RP of article.
The form for the relevant data source of commodity that the acquiring unit obtains are as follows:
P=Ui | i=1,2,3 ...
Q=Vj | j=1,2,3 ...
R=Rij | i ∈ U, j ∈ V }
S=Sij | i, j ∈ U }
RP=RPij | i ∈ U, j ∈ V }
Wherein: P is user's collection, and Ui is the user in user's set;Q is article collection, and Vj is the article in article set;R
For rating matrix, Rij indicates user i to the score value of article j;S is social matrix, and Sij is that user i closes the trust of user j
System;RP is preference matrix, and RPij indicates the score value that user i predicts article j.
In computing unit, the calculation difference of trust strength TN, similarity TS, comprehensive degree of belief T between user u, v
It is as follows:
Tu,v=α × TNu,v+(1-α)×TSu,v
In above formula, TNu,vIndicate trust strength of the user u to user v, d-(nv) indicate the quantity that user v is concerned, d+
(nu) indicate the number of users that user u is paid close attention to;TSu,vIndicate similarity of the user u to user v, LuIndicate the trust good friend of user u
List collection, LvIndicate the trust buddy list set of user v;Tu,vUser u is indicated to the synthesis degree of belief of user v, α is indicated
Award coefficient of reliability.
Rating matrix R, comprehensive trust degree matrix T, comment document sets D are inputted as follows based on the row of Multi-source Information Fusion
Sequence proposed algorithm model L (u, v), is trained, and when function penalty values are less than certain value, obtains the hiding attributive character of user
Matrix U, the hiding attributive character V of article:
In above formula, M indicates number of users;N indicates number of articles;U is that user hides attributive character, and V is that article hides category
Property feature;λrel、λrev、λwFor regularization coefficient, in order to prevent model over-fitting;Cnn is to extract article text information
The convolutional neural networks model of feature is hidden, W is convolutional neural networks parameter matrix, and X is the document input vector of model.
The matrix U and V obtained according to training calculates user to the preference matrix RP of article, and calculation is as follows:
RP=UTV
The merchandise display module 2 includes display unit and feedback unit;The display unit is used for from user to article
Preference matrix in obtain K commodity before preference value comes, shown in the page;The feedback unit is newest for collecting
Interbehavior data of the user to commodity, scoring and comment such as user to commodity, there are also information such as good friends of user's concern,
And these data are analyzed, weight is adjusted in real time.
The adjustment module 3 is for adjusting rating matrix, trusting matrix, comment document sets and preference matrix.
Claims (5)
1. a kind of intelligence commercial product recommending system, including sequentially connected information initializing module (1), merchandise display module (2) and
It adjusts module (3), which is characterized in that the information initializing module (1) is used to construct rating matrix, the user of user and article
The comment document sets for trusting matrix, consumer articles between user, preference matrix of the initialising subscriber to article;The commodity
Display module (2) current scene for identification is calculated before preference value comes K by preference matrix of the user to article
Commodity, and merchandise news is shown in the software page, and collect user to the feedback information of commodity;The adjustment module (3) is used
In adjustment rating matrix, trust matrix, comment document sets and preference matrix.
2. intelligence commercial product recommending system according to claim 1, which is characterized in that information initializing module (1) packet
Acquiring unit and computing unit are included, the acquiring unit is for obtaining the relevant data source of commodity;The computing unit is based on
Trust strength TN [u, v], similarity TS [u, v], comprehensive degree of belief T [u, the v] between user u and user v are calculated, and calculates update
Preference matrix RP of the user to article.
3. intelligence commercial product recommending system according to claim 2, which is characterized in that the commodity phase that the acquiring unit obtains
The form of the data source of pass are as follows:
P=Ui | i=1,2,3 ...
Q=Vj | j=1,2,3 ...
R=Rij | i ∈ U, j ∈ V }
S=Sij | i, j ∈ U }
RP=RPij | i ∈ U, j ∈ V }
Wherein: P is user's collection, and Ui is the user in user's set;Q is article collection, and Vj is the article in article set;R is to comment
Sub-matrix, Rij indicate user i to the score value of article j;S is social matrix, and Sij is trusting relationship of the user i to user j;RP
It is preference matrix, RPij indicates the score value that user i predicts article j.
4. it is according to claim 2 intelligence commercial product recommending system, which is characterized in that in the computing unit, user u, v it
Between trust strength TN, similarity TS, the calculation difference of comprehensive degree of belief T it is as follows:
Tu,v=α × TNu,v+(1-α)×TSu,v;
In above formula, TNu,vIndicate trust strength of the user u to user v, d-(nv) indicate the quantity that user v is concerned, d+(nu) table
Show the number of users of user u concern;TSu,vIndicate similarity of the user u to user v, LuIndicate the trust buddy list of user u
Set, LvIndicate the trust buddy list set of user v;Tu,vUser u is indicated to the synthesis degree of belief of user v, α indicates credit
Spend coefficient;
By rating matrix R, comprehensive trust degree matrix T, comment document sets D, the sequence based on Multi-source Information Fusion is pushed away as follows for input
Algorithm model L (u, v) is recommended, is trained, when function penalty values are less than certain value, obtains the hiding attributive character matrix of user
U, the hiding attributive character V of article:
In above formula, M indicates number of users;N indicates number of articles;U is that user hides attributive character, and V is that article hides attribute spy
Sign;λrel、λrev、λwFor regularization coefficient, in order to prevent model over-fitting;Cnn is to extract article Text information hiding
The convolutional neural networks model of feature, W are convolutional neural networks parameter matrix, and X is the document input vector of model;
The matrix U and V obtained according to training calculates user to the preference matrix RP of article, and calculation is as follows:
RP=UTV。
5. intelligence commercial product recommending system according to claim 1, which is characterized in that the merchandise display module (2) includes
Display unit and feedback unit;The display unit comes preceding K for obtaining preference value in the preference matrix from user to article
The commodity of position, are shown in the page;The feedback unit is used to collect newest user to the interbehavior data of commodity, and
These data are analyzed, adjust weight in real time.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910089640.7A CN109886779A (en) | 2019-01-30 | 2019-01-30 | A kind of intelligence commercial product recommending system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910089640.7A CN109886779A (en) | 2019-01-30 | 2019-01-30 | A kind of intelligence commercial product recommending system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109886779A true CN109886779A (en) | 2019-06-14 |
Family
ID=66927408
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910089640.7A Pending CN109886779A (en) | 2019-01-30 | 2019-01-30 | A kind of intelligence commercial product recommending system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109886779A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110866637A (en) * | 2019-11-06 | 2020-03-06 | 湖南大学 | Scoring prediction method, scoring prediction device, computer equipment and storage medium |
CN111127146A (en) * | 2019-12-19 | 2020-05-08 | 江西财经大学 | Information recommendation method and system based on convolutional neural network and noise reduction self-encoder |
CN112581153A (en) * | 2019-09-29 | 2021-03-30 | 拉扎斯网络科技(上海)有限公司 | Resource allocation method, resource allocation device, storage medium, and electronic apparatus |
CN112883289A (en) * | 2021-04-16 | 2021-06-01 | 河北工程大学 | PMF recommendation method based on social trust and tag semantic similarity |
CN113204713A (en) * | 2021-07-05 | 2021-08-03 | 南京信息工程大学 | Core user and core article mining method in large-scale commodity sales |
WO2022021383A1 (en) * | 2020-07-31 | 2022-02-03 | 深圳齐心集团股份有限公司 | Stationery recommendation system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20140087065A (en) * | 2012-12-21 | 2014-07-09 | 서울대학교산학협력단 | Method and system for Social Recommendation with Link Prediction |
CN105740430A (en) * | 2016-01-29 | 2016-07-06 | 大连理工大学 | Personalized recommendation method with socialization information fused |
CN108287904A (en) * | 2018-05-09 | 2018-07-17 | 重庆邮电大学 | A kind of document context perception recommendation method decomposed based on socialization convolution matrix |
CN108470075A (en) * | 2018-04-12 | 2018-08-31 | 重庆邮电大学 | A kind of socialization recommendation method of sequencing-oriented prediction |
-
2019
- 2019-01-30 CN CN201910089640.7A patent/CN109886779A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20140087065A (en) * | 2012-12-21 | 2014-07-09 | 서울대학교산학협력단 | Method and system for Social Recommendation with Link Prediction |
CN105740430A (en) * | 2016-01-29 | 2016-07-06 | 大连理工大学 | Personalized recommendation method with socialization information fused |
CN108470075A (en) * | 2018-04-12 | 2018-08-31 | 重庆邮电大学 | A kind of socialization recommendation method of sequencing-oriented prediction |
CN108287904A (en) * | 2018-05-09 | 2018-07-17 | 重庆邮电大学 | A kind of document context perception recommendation method decomposed based on socialization convolution matrix |
Non-Patent Citations (1)
Title |
---|
徐亚元: "一种融合了社交信息和文本信息的排序推荐算法", 《工业控制计算机》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112581153A (en) * | 2019-09-29 | 2021-03-30 | 拉扎斯网络科技(上海)有限公司 | Resource allocation method, resource allocation device, storage medium, and electronic apparatus |
CN112581153B (en) * | 2019-09-29 | 2023-04-07 | 拉扎斯网络科技(上海)有限公司 | Resource allocation method, resource allocation device, storage medium, and electronic apparatus |
CN110866637A (en) * | 2019-11-06 | 2020-03-06 | 湖南大学 | Scoring prediction method, scoring prediction device, computer equipment and storage medium |
CN110866637B (en) * | 2019-11-06 | 2022-07-05 | 湖南大学 | Scoring prediction method, scoring prediction device, computer equipment and storage medium |
CN111127146A (en) * | 2019-12-19 | 2020-05-08 | 江西财经大学 | Information recommendation method and system based on convolutional neural network and noise reduction self-encoder |
CN111127146B (en) * | 2019-12-19 | 2023-05-26 | 江西财经大学 | Information recommendation method and system based on convolutional neural network and noise reduction self-encoder |
WO2022021383A1 (en) * | 2020-07-31 | 2022-02-03 | 深圳齐心集团股份有限公司 | Stationery recommendation system |
CN112883289A (en) * | 2021-04-16 | 2021-06-01 | 河北工程大学 | PMF recommendation method based on social trust and tag semantic similarity |
CN112883289B (en) * | 2021-04-16 | 2022-05-06 | 河北工程大学 | PMF recommendation method based on social trust and tag semantic similarity |
CN113204713A (en) * | 2021-07-05 | 2021-08-03 | 南京信息工程大学 | Core user and core article mining method in large-scale commodity sales |
CN113204713B (en) * | 2021-07-05 | 2021-10-08 | 南京信息工程大学 | Core user and core article mining method in large-scale commodity sales |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109886779A (en) | A kind of intelligence commercial product recommending system | |
Cheng et al. | Aspect-aware latent factor model: Rating prediction with ratings and reviews | |
CN106802956B (en) | Movie recommendation method based on weighted heterogeneous information network | |
CN104063481B (en) | A kind of film personalized recommendation method based on the real-time interest vector of user | |
CN104935963B (en) | A kind of video recommendation method based on timing driving | |
CN103544216B (en) | The information recommendation method and system of a kind of combination picture material and keyword | |
CN110275964A (en) | The recommended models of knowledge based map and Recognition with Recurrent Neural Network | |
Cui et al. | A novel context-aware recommendation algorithm with two-level SVD in social networks | |
CN107220365A (en) | Accurate commending system and method based on collaborative filtering and correlation rule parallel processing | |
CN110442781A (en) | It is a kind of based on generate confrontation network to grade ranked items recommended method | |
CN106104512A (en) | System and method for active obtaining social data | |
CN109241440A (en) | It is a kind of based on deep learning towards implicit feedback recommended method | |
CN108363804A (en) | Local model weighted fusion Top-N movie recommendation method based on user clustering | |
CN102968506A (en) | Personalized collaborative filtering recommendation method based on extension characteristic vectors | |
CN106897914A (en) | A kind of Method of Commodity Recommendation and system based on topic model | |
CN102231166A (en) | Collaborative recommendation method based on social context | |
CN111915409B (en) | Item recommending method, device, equipment and storage medium based on item | |
Rakesh et al. | Linked variational autoencoders for inferring substitutable and supplementary items | |
Li et al. | Content-based filtering recommendation algorithm using HMM | |
CN109190030A (en) | Merge the implicit feedback recommended method of node2vec and deep neural network | |
CN109933721A (en) | A kind of interpretable recommended method merging user concealed article preference and implicit trust | |
CN106157156A (en) | A kind of cooperation recommending system based on communities of users | |
Khatter et al. | Movie recommendation system using cosine similarity with sentiment analysis | |
CN110083764A (en) | A kind of collaborative filtering cold start-up way to solve the problem | |
CN109272390A (en) | The personalized recommendation method of fusion scoring and label information |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190614 |
|
RJ01 | Rejection of invention patent application after publication |