A personalized recommendation system based on product attribute-specific weights and improved user behavior analysis
Jehwan Oh, Ok‐Ran Jeong, Eunseok Lee · 2010
The amount of information on the Web is increasing exponentially due to the rapid development of information communication technology. In the field of e-business in particular, research into recommendation systems that analyze user preferences and suggest related products has been an issue. Analyzing the preferences of users is very important for making recommendations. However, if there is a lack of user information when a predictive method is being applied, no suggestion can be made. Therefore, this paper proposes a personalized recommendation system that utilizes the information derived from analyzing the data regarding user behaviors. The proposed system observes the various actions of a user on the Web in order to accurately understand his or her intentions. In addition, the most preferred qualities of various products is inferred based on the ID3 algorithm and the Naive Bayesian algorithm is applied to analyze user preferences by giving different weight to different product attributes. In this way, the proposed system can provide accurate recommendations even in situations where data regarding the behavior of a user is lacking. In order to evaluate the efficiency of the proposed system, the recommendation results and the number of available behavior data were compared, thus confirming its effectiveness.