The Research Of Electronic Commerce Recommendation Algorithm Based On The Collaborative Filtering

With the development of science and technology, the electronic commerce becomes a necessary part of our life. The public enjoy the convenience of electronic commerce, however, they fall into the trouble of information overload. It is difficult for users to search useful information from the massive information. Thus, it is high time to develop a tool to overcome this problem. The personalized recommendation system solves the problem some extent.The collaborative filtering recommendation algorithm is one of the most widely used recommendation technology in many e-commerce personalized recommendation algorithms. However, the traditional collaborative filtering algorithm exists a lot of problems and has a big improvement of performance. With the progress of time, the number of users and items is growing in the recommendation system and the user-item matrix becomes sparse more and more. The similarity of users and items which calculated from the user-item matrix alone will have a limit precision. Therefore, if we just use the user’s rating matrix to recommend the items for users, the performance of recommendation system will degrade. In order to solve this problem, a novel collaborative filtering recommendation algorithm which combines the similarity and the credibility is proposed. At first, the rating of the users is exploited to calculate the original similarity. Then, the user-item matrix is used to calculate the credibility of the users. Finally, we combine the similarity with the credibility by linear combination to form a new recommendation weight. Comparing the other collaborative filtering recommendation algorithms which only use the user and item matrix to calculate the similarity, our method has a dense recommendation matrix and can get more neighbors.What’s more, the current recommendation depends on the user’s rating completely, while ignores the user’s emotional criticism of the product. In this paper, we use the cloud model based on the emotional tendency comparison algorithm to extract the user’s comments on the emotional word and transform the emotional word to a number. Then these numbers are used to construct the similarity of emotional. The experiments show that the collaborative filtering algorithm which used the similarity of emotional has an accurate rate of recommendation some extent. Furthermore, if the users have an obvious emotional tendency, the algorithm has a better performance of recommendation.