The Research of Modified Collaborative Filtering Recommendation Algorithm

Collaborative filtering is a kind of data mining algorithm. It has been widely used in commercial recommender system. In this paper, we discuss the problems of traditional collaborative filtering algorithm and propose improvement ideas. As the judging criteria of the recommendation algorithm, we introduce Mean Absolute Error (MAE). On the basis of data mining platform, we build user application layer of the prototype system by PHP language. We use the MovieLens data sets to conduct comparisons with the traditional and the improved collaborative filtering algorithm which is based on the item, then obtain different experimental data from two aspects, and evaluate the effect of the recommended result by MAE. Finally it concluded that the improved collaborative filtering recommendation algorithm is slightly better than the traditional collaborative filtering algorithm in the same situation.