Research On E-Commerce Personalized Recommendation System Based On Mobile Platform

With the advent of the Internet Era, e-commerce with its affordable, convenient shopping advantages gradually won the favor of people, which made online shopping become a widespread consumer behavior. However, with the expansion of commercial scale and commodity quantity, e-commerce has presented the trend of information overload. Huge goods information brings in more choices for users, which also makes users timely search for their satisfactory commodities inconveniently. The burden is more outstanding under the impact of mobile Internet. The development of mobile Internet technology makes mobile phones and other mobile devices gradually become the main way to access to information by replacing the traditional PC. The rapid growth of mobile network user scale promotes e-commerce extend from PC to mobile field, and opens a new business model of mobile e-commerce. Due to the limitation of display and processing capacity of mobile terminal, mobile e-commerce will face a more serious problem of information overload. Personalized recommendation system can increase customer satisfaction and loyalty and promote the sale of products through real-time and accurate personalized recommendation service for users. Therefore, in order to make the mobile terminal users find goods quickly and easily, the research on personalized recommendation system based on mobile platform is very important.Firstly, basic theories of e-commerce personalized recommendation system are studied on aspects of concept, function and overall architecture. And CF algorithm is found that has obvious advantages than other personalized recommendation algorithms. Then this paper does some research on CF algorithm. First, this paper describes the concept, classification and bottleneck problem of traditional CF algorithm. Second, this paper chooses slope one algorithm that is more efficient and accurate than traditional CF algorithm for research and analyses the advantages, principle and process of it. Shortcomings of slope one algorithm are pointed out that it doesn’t consider user interest changes and user similarity. Third, this paper puts forward improved scheme of slope one algorithm based on user interest forgetting function and user nearest neighbors, proves the feasibility of improved scheme by experimental test on MovieLens dataset. The improved algorithm is putted into practice in this paper by using android as the base platform of research and elaborating the design and implementation of personalized recommendation system based on android platform. Finally the research work is summarized and prospected in this paper.At present, studies of e-commerce personalized recommendation system based on mobile platform have just started. The research of this paper has a certain reference value for e-commerce enterprises to build personalized recommendation system on mobile platform.