Personalized Recommendation Collaborative Filtering Algorithm Based On Multi-user Interest

With the growing surge of the popularity of the network, the increasingly sharp increase of the amount of information on the Internet brings the problem of information overload. Personalized recommendation system has become an important component of e-commerce, as a technology to solve the "information overload" problem. In recent years, personalized recommendation system has been greatly developed in theoretical research and practical application. Almost all business sites use a personalized recommendation system to provide users with personalized service, and to enjoy a better user experience.Collaborative filtering personalized recommendation is one of the most important technologies in the field of recommendation systems. The goal of collaborative filtering personalized recommendation is to provide users with a more accurate list of items. Common sense tells us that the user’s interest is often diversity and preference is different for different types of goods. Existing collaborative filtering personalized recommendation has no good way to solve this problem. The study in this paper is how to provide users with a personalized recommendation service accurately that suit their interests in the diversity of user interest.The research content and innovative work of this paper is mainly reflected in the following aspects:(1) Carrying out in-depth research and analysis for existing collaborative filtering personalized recommendation algorithm, pointing that user’s multi-interest problems affect of collaborative filtering personalized recommendation algorithm.(2) For users and more interest, this paper proposes a novel similarity calculation method. Weight coefficient8balances proportion between the similarity between the project and the user’s interest. By appropriate adjustment of the value of k, the nearest neighbor that is calculated is more in line with certain types of user’s interest. In the calculation of the prediction score, weighted average to account with the weight of the user interest as coefficient, prediction score is more accurate.(3) For the data sparseness problem in the certain type that is caused by user’s multiple interests, this paper propose a prediction score on the target project by user-based collaborative filtering personalized recommendation, this improves the data sparseness problem. Then the user-based collaborative filtering techniques is applied to project-based collaborative filtering technology, to avoid the drawbacks of potential interest can’t be recommended in item-based recommendation algorithm.In Matlab environment, this paper carries out a simulation experiment for improved algorithm for personalized recommendation based on the user’s multiple interests collaborative filtering. And taking into account the time complexity, this paper take the average running time (MRT) as a new evaluation standards. By Comparison with the existing algorithms, verifying the improved algorithm is proposed in the paper better than other algorithms in terms of recommendation accuracy and real-time, thus help to provide a more personalized recommendations.