The following information was submitted:
Transactions: WSEAS TRANSACTIONS ON INFORMATION SCIENCE AND APPLICATIONS
Transactions ID Number: 29-710
Full Name: Tsuyoshi Takayama
Position: Associate Professor
Age: ON
Sex: Male
Address: Graduate School of Software and Information Science, Iwate Prefectural University, 152-52, Sugo, Takizawa, Iwate020-0193
Country: JAPAN
Tel: +81-19-694-2614
Tel prefix:
Fax: +81-19-694-2501
E-mail address: takayama@soft.iwate-pu.ac.jp
Other E-mails: takayama@iwate-pu.ac.jp
Title of the Paper: Recommendation Method That Considers the Context of Product Purchases
Authors as they appear in the Paper:
Email addresses of all the authors:
Number of paper pages: 10
Abstract: We propose herein a technique for product recommendation in E-commerce by considering the context of product purchases, and verify the effectiveness of the technique through an evaluation experiment. Researchers have been aggressively studying techniques that can be used by stores to recommend to customers products that have relatively high purchase potential. Collaborative filtering is representative of conventional techniques. However, the collaborative filtering technique is based on the hypothesis that similar customers purchase similar products, and the context of product purchases is not considered in full. In the present study, a context matrix by which to manage the context history of product purchases is proposed. Collaborative filtering cannot distinguish the following two facts that `Product B was purchased after Product Af and `Product A was purchased after Product Bf. The context matrix, however, enables such information to be expressed and manag!
ed separately. We also propose four types of context matrix update methods which differs in subset selection on purchase history and user selection on obtaining purchase history. The results of an evaluation experiment revealed the following: i) The proposed technique can improve the recommendation precision by taking into account the context of purchases when making recommendations. ii) As the amount of available purchase history and context data increases, the recommendation precision improves. iii) The highest recommendation precision among four types of context matrix update methods is obtained, if all contexts of purchases along time axis by customers of similar taste only are considered.
Keywords: Recommendation, database, data mining, and E-commerce.
EXTENSION of the file: .pdf
Special (Invited) Session: Proposal and Evaluation of a Recommendation Technique That Considers the Context of Product Purchases
Organizer of the Session: 618-299
How Did you learn about congress:
IP ADDRESS: 210.156.40.65
