Research On The Improved Collaborative Filtering Algorithm In Recommendation System

With the rapid development of network technology, the Internet provides users with more and more information, information overload is becoming an extremely serious problem. How to recommend quickly the information to the users by their preference from a lot of information is the concern to the Internet operators and many scholars in recent years, in this case, the recommendation system came into being.As the core of the recommendation system, recommendation algorithm is also of concern to more and more scholars. Among all the algorithms, the most extensive and most recognized method is the Collaborative Filtering algorithm. Based on the research and analysis of Collaborative Filtering algorithm, the article summarized the deficiencie(sCold Start and Sparse Data)of the existing collaborative filtering algorithm and its improved algorithm. For cold start problems, the article proposes to calculate the items' similarity by their multi-properties and multi- categories. By this way, we can calculate the items' similarity without the users'score, and solve the cold start problem; meanwhile, the paper uses a clustering method before using the improved collaborative filtering to set the user-items data into several categories. Firstly, calculating the similarity between the candidate items and target items in each category, and finally calculating the prediction score unify. The advantage is that the algorithm can improve the timeliness and recommend quality, and reduce the data sparseness; at the last of the article, we use the data which from the Movie Lens where is providing the films'user-item data to do the algorithm simulation experiment. The experiment shows that the improved algorithm makes a higher quality recommendation results than the traditional Collaborative Filtering algorithms.