考慮使用者喜好變化：一種基於行為序列的推薦方法

• Main Authors: Li-Wei Chao, 趙力緯
• Other Authors: Yen-Hsien Lee
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/ca7a76

碩士 === 國立嘉義大學 === 資訊管理學系研究所 === 104 === The rapid development of network technology has boomed the E-commerce web sites that facilitate people shopping online. However, the volumes and great diversity of products available in online stores nowadays has caused the problem of information overload. To alleviate customers’ information loading and to enhance their shopping experiences, lots of E-commerce web sites developed product recommender systems to provide personalized product recommendation services for their target customers. Collaborative Filtering (CF), also known as social filtering or user-to-user correlation analysis, has been one of the promising recommendation methods for product recommender system. CF exploits known user preferences to associate a customer with referent others exhibiting similar preferences, and then leverage their preferences to recommend products to that customer. Though CF seems to be effective in recommendations, it takes customer’s preferences as static. Specifically, it analyzes customers’ preferences as a set of timeless records and neglects the time dimension existing in customers’ behavior histories. Considering customer’s behavioural sequence shall be important in making recommendations, because it provides a description of evolution and/or changes in his/her preferences over time. In this study, we intended to propose a revised CF method that could consider the target customer’s purchase sequence when assessing the neighbourhood and making recommendations. We empirically evaluated the performance of the proposed method using the MovieLens dataset. Our evaluation results show that the proposed method outperforms the benchmark (i.e., traditional CF method) in almost all evaluation criteria. The results also suggest the consideration of customer’s behavioural sequence is important to recommender systems when making appropriate recommendations.