CN1588400A - Method for recommending goods in electronic business - Google Patents

Method for recommending goods in electronic business Download PDF

Info

Publication number
CN1588400A
CN1588400A CN 200410067331 CN200410067331A CN1588400A CN 1588400 A CN1588400 A CN 1588400A CN 200410067331 CN200410067331 CN 200410067331 CN 200410067331 A CN200410067331 A CN 200410067331A CN 1588400 A CN1588400 A CN 1588400A
Authority
CN
China
Prior art keywords
commodity
client
scoring
related coefficient
same feeling
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
Application number
CN 200410067331
Other languages
Chinese (zh)
Inventor
申瑞民
谢波
韩鹏
杨帆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Jiaotong University
Original Assignee
Shanghai Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Jiaotong University filed Critical Shanghai Jiaotong University
Priority to CN 200410067331 priority Critical patent/CN1588400A/en
Publication of CN1588400A publication Critical patent/CN1588400A/en
Pending legal-status Critical Current

Links

Images

Abstract

The present invention relates to commodity recommending method in electronic business affairs, and belongs to the field of network technology. The present invention allows customers to browses commodity in network browser and input the evaluation on the commodity, the network server stores these customers' penchant information in its data base for creating the commodity recommendation. The present invention evaluates commodity with consensus correlation coefficient expressing the customer's evaluation, forecasts the evaluation of unknown commodity with the consensus correlation coefficient and known commodity evaluation, and recommends the commodity with high evaluation to customers. The present invention can process effectively massive customer and commodity data and promote the commodity selling of electronic business system.

Description

The method of Recommendations in the ecommerce
Technical field
The present invention relates to a kind of method of computer recommending commodity of networking technology area, specifically is the method for Recommendations in a kind of ecommerce.
Background technology
Along with popularizing and Development of E-business of internet; e-commerce system is when providing more and more the selection for the user; it is complicated more that its structure also becomes, and the user gets lost in a large amount of merchandise news spaces through regular meeting, can't find the commodity that oneself need smoothly.Direct and the user interactions of Technologies of Recommendation System in E-Commerce, the simulation store sales personnel provide commercial product recommending to the user, helps the user to find required commodity, thereby finish purchasing process smoothly.Under the competitive environment that is growing more intense, Technologies of Recommendation System in E-Commerce can effectively keep the user, prevent customer loss, improves the sale of e-commerce system.
Find by prior art documents, Sarwar is at " Item-based collaborativeFiltering recommendation algorithms (project-based collaborative filtering recommending algorithm) " (InProceedings of the Tenth International World Wide Web Conference, 2001, Pages.285-295. " the 10th international WWW conference collection of thesis ", calendar year 2001, the 285-295 page or leaf) in this piece article, a kind of project-based collaborative filtering recommending algorithm has been proposed.In this piece paper, Sarwar proposes to use Pearson's related coefficient (formula 1) to weigh the degree of correlation between the commodity, by client known commodity hobby scoring and the related coefficient between the commodity are calculated the prediction of client to unknown commodity fancy grade, thereby the most possible interested merchant's crystalline substance of client is recommended client.Because Pearson's related coefficient can be analyzed the degree of correlation between the commodity better, therefore existing Technologies of Recommendation System in E-Commerce has been continued to use this method mostly.
Figure A20041006733100042
Related coefficient between expression commodity i and the commodity j; U represents simultaneously to give the client's of commodity i and commodity j scoring set; v U, iRepresent the scoring of client u to commodity i, The average score of expression client u.
But be to use Pearson's related coefficient to measure degree of correlation between the commodity too big problem of the amount of recomputating can bring Data Update the time.For example, certain client increases a hobby scoring to certain commodity, Pearson's related coefficient between these commodity and the every other commodity all needs to recomputate, need recomputate Pearson's related coefficient between the inferior commodity of O (N) for a Technologies of Recommendation System in E-Commerce that has M client and N spare commodity, complexity computing time of each Pearson's related coefficient is O (M), and therefore whole time complexity of reruning is O (M*N).The time complexity O (M*N) that reruns under the few situation of client and commodity number is also little, does not have what problem; But when meeting with mass users and commodity number, Technologies of Recommendation System in E-Commerce will be difficult to generation in real time because complexity O computing time (M*N) is too big and recommend even can't operate as normal.
Summary of the invention
The objective of the invention is to overcome deficiency of the prior art, a kind of method that is used for ecommerce to the customer recommendation commodity is proposed, make client can in the web browser picture, browse commodity and the own evaluation of input to commodity, the webserver obtains these client's preference information and deposits it in database, and from this database, reason out client and may like what commodity, output it to client for your guidance.The webserver does not use Pearson's related coefficient, and the client's number that two commodity is had identical scoring is defined as two same feeling related coefficients between the commodity, and replace Pearson's related coefficient to weigh degree of correlation between the commodity with same feeling related coefficient, so each client increases new hobby scoring to certain commodity does not just need to recomputate the Pearson's related coefficient between these commodity and the every other commodity, only need the simple commodity same feeling correlation matrix of adjusting, required time complexity is that (K is the commodity number that each client's average score is crossed in the system to O (K), K＜＜N), time complexity is reduced to O (K) from O (M*N), and has improved the accuracy rate of Technologies of Recommendation System in E-Commerce.
The present invention is achieved by the following technical solutions, the present invention browses commodity by client in the web browser picture, and input is to the evaluation of commodity, the webserver obtains these client's preference information and deposits it in database, generation is to client's commercial product recommending, promptly the client's number that two commodity is had identical scoring is defined as two same feeling related coefficients between the commodity, and weigh degree of correlation between the commodity with same feeling related coefficient, mark according to same feeling related coefficient and known commodity and to predict the scoring of client to unknown commodity, the commercial product recommending that scoring is high is given client.
Same feeling related coefficient between the described commodity calculates by following formula:
The client's number that two commodity is had identical hobby is many more, and then these two commodity are relevant more.Use formula (3) calculates the same feeling related coefficient between the commodity.
Figure A20041006733100061
D i , j = Σ k = 1 M d ( i , j , k ) - - ( 3 )
The implication of each variable is as follows in the formula:
D I, jRepresent the same feeling related coefficient between commodity i and the commodity j
M represents the client's number in the system
(k) whether expression client k has identical hobby to commodity i with commodity j to d for i, j
v K, iExpression client k is to the scoring of commodity i.
The client's number that two commodity is had identical scoring is defined as two same feeling related coefficients between the commodity, and weigh degree of correlation between the commodity with same feeling related coefficient, for each newly-increased hobby scoring, do not need again whole calculating commodity correlation matrix, only need be according to the corresponding adjustment commodity of following steps correlation matrix: establishing client U increases new hobby scoring V to commodity I, other hobby scorings for client U also are each commodity J of V, because client U has identical hobby to commodity J and I, so the same feeling related coefficient between commodity J and the I adds up one.
The predicted value of described unknown commodity scoring obtains by following formula:
Use formula (4) to predict the fancy grade of client to unknown commodity.
P a , j = Σ r ∈ R ( D j , r * v a , r ) Σ r ∈ R ( | D j , r | ) - - ( 4 )
The implication of each variable is as follows in the formula:
P A, jExpression client a is to the scoring predicted value of unknown commodity j
D J, rRepresent the same feeling related coefficient between unknown commodity j and the known commodity r
v A, rExpression client a is to the scoring of known commodity r
R represents among the client a and the set of all known commodity that unknown commodity j same feeling related coefficient is higher.
The present invention can handle magnanimity client and commodity data effectively, keeps the user, prevents customer loss, improves the merchandise sales of e-commerce system.Experimental result shows, adopt the present invention after owing to do not need, so time complexity is reduced to O (K) from O (M*N) because the Pearson's related coefficient between all commodity is recomputated in newly-increased hobby scoring.And the present invention defines and has used same feeling related coefficient, replace Pearson's related coefficient to weigh degree of correlation between the commodity with it, owing to rejected " noise " data of using Pearson's related coefficient to introduce, therefore made recommendation accuracy of the present invention be higher than the method for common use Pearson related coefficient.
Description of drawings
Fig. 1 is the inventive method process flow diagram.
Fig. 2 likes the scoring synoptic diagram for beginning client at the beginning of the embodiment of the invention.
Fig. 3 likes the scoring synoptic diagram for the newly-increased client of the embodiment of the invention.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, below in conjunction with embodiment and accompanying drawing, the present invention is described in more detail.
At first, as shown in Figure 1, client can browse commodity in the web browser picture, and the own evaluation of input to commodity, the webserver is liked score data with these clients and is deposited database in, and uses " calculating the correlativitys between the commodity " this program module to calculate same feeling related coefficient between the commodity according to formula 3.For example, as shown in Figure 2, hobby scoring scope is 1 to 5 minute, not scoring of 0 expression, and client A, B, C are respectively (0,4,4) to the hobby of commodity E, F, G, (1,3,0) and (1,1,4).Result of calculation is as shown in table 1: having a client C identical with the scoring of F hobby to commodity E, all is 1 minute, and then the same feeling related coefficient of commodity E and F is 1; Do not have client identical to the hobby scoring of commodity E and G, then the same feeling related coefficient of commodity E and G is 0; Having a client A identical with the G scoring to commodity F, all is 4 minutes, and then the same feeling related coefficient of commodity F and G is 1.
Table 1 is the same feeling related coefficient between the beginning commodity at the beginning of the embodiment of the invention:
Same feeling related coefficient ????E ????F ????G
?????E ????N/A ????1 ????0
?????F ????1 ????N/A ????1
?????G ????0 ????1 ????N/A
The second, when client increases a scoring newly, as shown in Figure 3, the newly-increased scoring to commodity E of client A is 4 minutes.The same feeling related coefficient result of calculation of this moment is as shown in table 2, because being 4 minutes commodity, client A scoring also has F and G, therefore need be commodity E and F, and the same feeling related coefficient between commodity E and the G adds up one respectively, the same feeling related coefficient of commodity E and F becomes 2 from 1 like this; The same feeling related coefficient of commodity E and G is 1 from 0; It is 1 constant that the same feeling related coefficient of commodity F and G keeps.
Table 2 is liked the same feeling related coefficient of marking between the commodity of back for the newly-increased client of the embodiment of the invention:
Same feeling related coefficient ???E ????F ????G
??????E ???N/A ????2 ????1
??????F ???2 ????N/A ????1
??????G ???1 ????1 ????N/A
The 3rd, the webserver uses " predictive user is to the hobby of unknown commodity " this program module after the same feeling related coefficient that obtains between the commodity, uses formula 4 to predict the hobby scoring of client to unknown commodity.As shown in Figure 3, known client B is 1 to the scoring of E, and the same feeling related coefficient of commodity E and G is 1, and client B is 3 to the scoring of commodity F, and the related coefficient of commodity F and G is 1, predicts the scoring of client B to unknown commodity G according to formula 4, P B , G = ( 1 * 1 + 1 * 3 ) ( 1 + 1 ) = 2 , Therefore client B is 2 minutes to the prediction scoring of unknown commodity G.
At last, the webserver uses " producing commercial product recommending for client's reference " this program module to generate commercial product recommending, in the above embodiments, hobby scoring scope is 1 to 5 minute, these commodity are liked in the high more expression of mark more, and client B is less than normal to the prediction scoring (2 minutes) of unknown commodity G, therefore not to client B Recommendations G.
The present invention can handle magnanimity client and commodity data effectively, keeps the user, prevents customer loss, improves the merchandise sales of e-commerce system.The EachMovie data set that provides of the HP Lab of Shi Yonging is as test data set with use mean absolute deviation to experimentize as the prediction accuracy criterion in accordance with international practices, the experimental result that obtains shows, the present invention's proposition and use same feeling related coefficient replacement Pearson related coefficient are weighed the degree of correlation between commodity, do not need to recomputate the Pearson's related coefficient between all commodity because of newly-increased hobby scoring, time complexity is reduced to O (K) from O (M*N), and removed and used Pearson's related coefficient to introduce " noise " data, made recommendation accuracy of the present invention be higher than the method 5% that tradition is used Pearson's related coefficient.

Claims (4)

1, the method of Recommendations in a kind of ecommerce, it is characterized in that, in the web browser picture, browse commodity by client, and input is to the evaluation of commodity, the webserver obtains these client's preference information and deposits it in database, generation is to client's commercial product recommending, promptly the client's number that two commodity is had identical scoring is defined as two same feeling related coefficients between the commodity, and weigh degree of correlation between the commodity with same feeling related coefficient, mark according to same feeling related coefficient and known commodity and to predict the scoring of client to unknown commodity, the commercial product recommending that scoring is high is given client.
2, the method for Recommendations in the ecommerce according to claim 1 is characterized in that, the same feeling related coefficient between the described commodity calculates by following formula:
The client's number that two commodity is had identical hobby is many more, and then these two commodity are relevant more, uses following formula to calculate same feeling related coefficient between the commodity,
D i , j = Σ k = 1 M d ( i , j , k )
The implication of each variable is as follows in the formula: D I, jRepresent the same feeling related coefficient between commodity i and the commodity j, M represents the client's number in the system, and (k) whether expression client k has identical hobby, v to commodity i with commodity j to d for i, j K, iExpression client k is to the scoring of commodity i.
3, method according to Recommendations in claim 1 or the 2 described ecommerce, it is characterized in that, the client's number that two commodity is had identical scoring is defined as two same feeling related coefficients between the commodity, and weigh degree of correlation between the commodity with same feeling related coefficient, for each newly-increased hobby scoring, according to the corresponding adjustment commodity of following steps correlation matrix: establishing client U increases new hobby scoring V to commodity I, other hobby scorings for client U also are each commodity J of V, because client U has identical hobby to commodity J and I, so the same feeling related coefficient between commodity J and the I adds up one.
4, the method for Recommendations in the ecommerce according to claim 1 is characterized in that, the predicted value of described unknown commodity scoring obtains by following formula:
P a , j = Σ r ∈ R ( D j , r * v a , r ) Σ r ∈ R ( | D j , r | )
The implication of each variable is as follows in the formula: P A, jExpression client a is to the scoring predicted value of unknown commodity j, D J, rRepresent the same feeling related coefficient between unknown commodity j and the known commodity r, v A, rExpression client a is to the scoring of known commodity r, and R represents among the client a and the set of all known commodity that unknown commodity j same feeling related coefficient is higher.
CN 200410067331 2004-10-21 2004-10-21 Method for recommending goods in electronic business Pending CN1588400A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200410067331 CN1588400A (en) 2004-10-21 2004-10-21 Method for recommending goods in electronic business

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200410067331 CN1588400A (en) 2004-10-21 2004-10-21 Method for recommending goods in electronic business

Publications (1)

Publication Number Publication Date
CN1588400A true CN1588400A (en) 2005-03-02

Family

ID=34604152

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200410067331 Pending CN1588400A (en) 2004-10-21 2004-10-21 Method for recommending goods in electronic business

Country Status (1)

Country Link
CN (1) CN1588400A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101206749A (en) * 2006-12-19 2008-06-25 株式会社G&G贸易公司 Merchandise recommending system and method thereof
WO2009079808A1 (en) * 2007-12-07 2009-07-02 Ebay Inc. System and method for creating social services based on buying experience
CN102789617A (en) * 2011-05-19 2012-11-21 乐活在线（北京）网络技术有限公司 Commodity information correlation method and system
CN103858142A (en) * 2011-09-29 2014-06-11 株式会社咕嘟妈咪 Store information provision system
CN104615631A (en) * 2014-10-29 2015-05-13 中国建设银行股份有限公司 Information recommendation method and device
CN106408483A (en) * 2016-08-31 2017-02-15 国信优易数据有限公司 Meteorology cloud intelligent business method and system
CN108205775A (en) * 2016-12-20 2018-06-26 阿里巴巴集团控股有限公司 The recommendation method, apparatus and client of a kind of business object
CN112734462A (en) * 2020-12-30 2021-04-30 北京字跳网络技术有限公司 Information recommendation method, device, equipment and medium

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101206749A (en) * 2006-12-19 2008-06-25 株式会社G&G贸易公司 Merchandise recommending system and method thereof
CN101206749B (en) * 2006-12-19 2013-06-05 株式会社G&G贸易公司 Merchandise recommending system and method using multi-path image retrieval module thereof
WO2009079808A1 (en) * 2007-12-07 2009-07-02 Ebay Inc. System and method for creating social services based on buying experience
CN102789617A (en) * 2011-05-19 2012-11-21 乐活在线（北京）网络技术有限公司 Commodity information correlation method and system
CN103858142A (en) * 2011-09-29 2014-06-11 株式会社咕嘟妈咪 Store information provision system
CN104615631A (en) * 2014-10-29 2015-05-13 中国建设银行股份有限公司 Information recommendation method and device
CN104615631B (en) * 2014-10-29 2019-02-12 中国建设银行股份有限公司 A kind of method and device of information recommendation
CN106408483A (en) * 2016-08-31 2017-02-15 国信优易数据有限公司 Meteorology cloud intelligent business method and system
CN108205775A (en) * 2016-12-20 2018-06-26 阿里巴巴集团控股有限公司 The recommendation method, apparatus and client of a kind of business object
CN112734462A (en) * 2020-12-30 2021-04-30 北京字跳网络技术有限公司 Information recommendation method, device, equipment and medium
CN112734462B (en) * 2020-12-30 2024-04-05 北京字跳网络技术有限公司 Information recommendation method, device, equipment and medium

Similar Documents

Publication Publication Date Title
Lu et al. BizSeeker: a hybrid semantic recommendation system for personalized government‐to‐business e‐services
Moe et al. Capturing evolving visit behavior in clickstream data
US20090125382A1 (en) Quantifying a Data Source's Reputation
US8103540B2 (en) System and method for influencing recommender system
US8751307B2 (en) Method for implementing online advertising
US7885849B2 (en) System and method for predicting demand for items
US7966342B2 (en) Method for monitoring link & content changes in web pages
Bauer et al. Recommender systems based on quantitative implicit customer feedback
US20150058331A1 (en) Search result ranking using machine learning
US20070005425A1 (en) Method and system for predicting consumer behavior
US20040249700A1 (en) System & method of identifying trendsetters
CN104063481A (en) Film individuation recommendation method based on user real-time interest vectors
CN102419779A (en) Method and device for personalized searching of commodities sequenced based on attributes
Stanujkic et al. An innovative approach to evaluation of the quality of websites in the tourism industry: A novel MCDM approach based on bipolar neutrosophic numbers and the hamming distance
JP2002157394A (en) Network marketing system
Ashenfelter et al. Empirical methods in merger analysis: Econometric analysis of pricing in FTC v. Staples
CN1588400A (en) Method for recommending goods in electronic business
Ginting et al. Integration of kansei engineering and quality function deployment (qfd) for product development: a literature review
Olson et al. Operational, economic and mission elements in not-for-profit organizations: the case of the Chicago Symphony Orchestra
Wang et al. Research on personalized recommendation based on web usage mining using collaborative filtering technique
Scuderi et al. ECONOMIC EVALUATION OF INNOVATIVE INVESTIMENTS IN AGRI-FOOD CHAIN.
Hooda et al. A study of recommender systems on social networks and content-based web systems
Lee et al. Determining service quality measurement key indicators in a travel website using a fuzzy analytic hierarchy process
Pratiwi et al. Investigating the Impact of Website Quality on User Satisfaction in IDN Times News Website
Ghose et al. Search Less, Find More? Examining Limited Consumer Search with Social Media and Product Search Engines.

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication