Research And Implementation Of Personalized Recommendation Algorithm For Mobile Client

Smart mobile devices is combined with traditional e-commerce greatly enhances shopping experience. However, smart mobile devices are limited in power, data processing, and network speed, which propels mobile e-commerce to ensure user satisfaction. Researches have shown that the existing personalized recommendation algorithms are generally facing the problems of data sparsity and cold start. Based on the “Supai e-commerce platform”, this paper discusses those problems and puts forward corresponding solutions. This paper deals with the following aspects:1. Putting forward a solution to the problem of cold start in mobile e-commerce platform. The solution consists of three steps:(1) extracting context features according to the clicking and browsing record of existing users;(2) training user interest classification model based on the context features;(3) predicting new users’ shopping interest through this model and completing the personalized recommendation.2. It proposes a new Collaborative Filtering Recommendation Algorithm by introducing users’ trust mechanism to solve the problem of recommendation accuracy and data sparsity. In this paper, the generation of users’ trust is divided into two parts — social reputation and social similarity. The process of model building is completed by extracting context features of user trust in the way of AHP. In order to match more neighbors to the users, the new algorithm combines the users’ most trusted neighbors and the users’ most similar neighbors in scoring in the way of threshold filtering. In addition, this paper also improves the existing scoring prediction formula to fully consider the influence of users’ trust relations on the recommendation results in the scoring prediction process. The feasibility and effectiveness of the improved algorithm are verified and proved by experiments.The result shows that the improved algorithm can effectively solve the problem of user cold start and significantly improve the quality of personalized recommendation system.