Modified collaborative filtering algorithm based on meta similarity
基于元相似度的推荐算法*

This paper presented a collaborative filtering algorithm based on a new similarity definition, namely meta similarity. By considering the vector correlation of the user similarity matrix, the meta similarity between any two users could be obtained. Even the relation of two users who had no common collected items could be investigated by it. Combining with the initial similarity with a tunable parameter, the integrated similarity could reflect relations between users more properly. Numerical results indicate that the algorithmic accuracy, measured by the average ranking score, and precision and recall is improved greatly.