E-commerce Recommendation System Based On Product Feature And Hierarchy

With the popularization of Internet and the development of the E-Commerce techniques, E-Commerce gives consumers more and more choice. At the same time, consumers often lose their ways when searching for products and can't find products quickly and exactly, because of the extension of information. By communicating with consumers directly and simulating the behavior of a salesman to suggest products for consumers, a recommendation system of E-Commerce could help consumers easier to find the products they really want to buy. In today's warm marketing competition, the recommendation system of E-Commerce could attract and retain consumers. So it can enhance the distribution and competition.Personalized recommend system has good application and foreground in E-Commerce. It becomes a main research issue of E-Commerce and attract more and more researchers' attention. Collaborative filtering technology is the main research issue in the field of personalized recommendation, and in this paper we focus our research on this issue.The paper first analyzes the problems in the traditional collaborative filtering arithmetic: As magnitudes and commodities grow rapidly, user ration data becomes extreme sparse, which results in the low quality of recommendation. To solute the promblem, a collaborative filtering arithmetic based on product hiberarchy and feature is proposed. When calculating the similarity of prodcts and the similarity of users,the arithmetic absorbs the idea of user-based and item-based collaborative filtering. Based on it, the concept of product hiberarchy and product feature is introduced,making the results more precise. When computing the initial forecast and final forecast rating of items, product classification is considered. This improves the veracity of the forecast results and the quality of recommendation. In order to validate the arithmetic, we use the date set of MovieLens, and compare with other arithmetics from different aspects. The results indicate that the proposed arithmetic reduces the sparsity of the user ration data. The proposed arithmetic is better than traditional collaborative filtering arithmetic in both initial forecast and final forecast rating of items.Further more, the proposed arithmetic is applied to an E-Commerce web set, and is made up considering the date character of the web set.