Study on internet recommendation system of collaborative filtering based on scatter difference

For the phonemon for over fitting in the current recommendation system, put forward using the fuzzy scatter difference to the recommendation algorithm. The algorithm avoids the robustness in original collaborative filtering recommendation system for only calculating the user´s similarities. The Experimental results show that the system has good accuracy and efficiency of the recommendation under the three indexes of MAE, RMSE and correlation, especially in the case of sparse initial data.