Multi-stage Collaborative Filtering Approach For Mobile Commerce

As the Internet application developing faster and faster, the web-based information is distributed, open and rapid accumulation. All of these cause the information overload and distribution of scattered and so on. It greatly affects the users accessing to information. In order to improve the user experience, the personalized recommendation service which based on users' characteristics and behavior developed gradually. The recommendation system, with its efficient, convenient features, has become an important channel to get information.In recent years, the development of wireless communication technology and mobile devices greatly changes people's work and life. Much attention has been devoted to Mobile Commerce because it can provide services anytime and anywhere.In the mobile commerce, the Location Bases Service is considered as a killer-class application. However, users still have to face the information overload in the Location Based Service. At the same time, the limitations of users' time, constantly changing environment and the weak display capability of the mobile terminals make the problem worse.In Electronic Commerce, the collaborative filtering recommendation system which based on users' interests and purchasing history and other information can solve the information overload effectively and reduce the burden of users. Because there are many differences between traditional Electronic Commerce and Mobile Commerce, the traditional collaborative recommendation system which is designed for internet users can't be directly transferred to the Mobile Commerce. First, location relative, emergency and access any time anywhere are the unique characters of Mobile Commerce; second, there are many differences between the two in technology, service characteristics and business models. Because of the small screen, weak processing capacity, the mobile recommendation service must meet some needs, such as recommendation information must be precise and concise, the result can be displayed in the limited screen. In the context of this, this paper studies the personalized recommendation system and its core technology, analyses the differences between Mobile Commerce and Electronic Commerce, and introduces the development and present condition of the Mobile Commerce system at home and aboard systematically. After researched the new features of collaborative filtering system in the Mobile Commerce, the paper proposed a multi-stage collaborative filtering algorithm, which consists of two stages. The first stage is based on user rating information collaborative filtering, and the second is based on the context information collaborative filtering. In the first stage, a two-phase clustering-based collaborative filtering algorithm processes the user rating information and item attribute to predict the rating. In the second stage, a Bayesian- network-based context information model deal with the context information of the recommended target generated in the first phase. After the information propagation, the output of the Bayesian network engine is an associated probability. Then we adjust the order of the recommendation items according to the probability. Collaborative filtering recommendation is the most successful personalized recommendation algorithm among current technologies. However, the data sparsity decreases the performance of recommendation. The paper suggests the two-phase clustering-based collaborative filtering algorithm. The algorithm not only reduces the sparsity of data and improves the accuracy of the nearest neighbor, but also improves the recommendation accuracy and reduces the time complexity compared with the traditional algorithms. Combined with the context information, the multi-stage collaborative filtering can improve the accuracy of recommendation, and meet the users'needs better.