Optimization And Application On Personalized Recommendation Algorithm Based On Collaborative Filtering

In the recent years, with the development and maturation of Internet and e-commerce, people get a lot of convenience in information acquisition and purchase of goods. At the same time, there is an "information overload" issue. Users waste a lot of time looking for satisfactory goods. Therefore, researchers pay more and more attention to the development of recommendation system. Recommendation system can provide users with the appropriate product recommendation based on the user's history behavior and personal information. It can help e-commerce systems improve the quality of service.Collaborative filtering is one of the successful applications of personalized recommendation technology in e-commerce recommendation system. It helps users quickly find products or information of interest by sharing the experience of other users. The research of collaborative filtering has become a hot field of personalized recommendation. This paper's major work is to improve collaborative filtering algorithms and make optimization. This paper describes some kinds of the recommendation techniques. Then focuses on the analysis of the current status of collaborative filtering algorithms, and does an in-depth research and analysis of various types of collaborative filtering algorithms to compare their respective characteristics.Main research of this paper is to improve and optimize the Slope one scheme. Slope one is an Item-based collaborative filtering recommendation algorithm. Slope one has the features of simple and efficient, but it ignores the impact of the number of evaluations. Through the analysis of its shortcomings, we discussed its two improved methods in detail:Weighted Slope one, R-Slope one. Although the Weighted Slope one and R-Slope one algorithms have a certain degree of improvement than the Slope one, but the two algorithms remain the user's specific problem and some other problems. Based on this, this paper presents two new algorithms which are based on the improved Slope one and user clustering:WSO-UC, RSO-UC.This paper designed two experiments to verify the performance of the proposed algorithms (WSO-UC and RSO-UC). One of the experiments is to compare WSO-UC and RSO-UC with Slope one, Weighted Slope one and R-Slope one. Another experiment is to compare WSO-UC and RSO-UC with several other commonly used collaborative filtering algorithms. Experimental results show that WSO-UC and RSO-UC have a better performance of forecast and recommendation. It proves the value of this paper's research. Finally, we describe the process that using WSO-UC and RSO-UC establish a recommendation model for e-commerce.