An Improved Collaborative Filtering Algorithm And Its Application In Recommendation System

The recommendation system is an advanced business intelligence platform based on massive data mining.He transforms the traditional system or platform to provide information to users in a passive way,and uses the recommendation algorithm to construct a user preference model.Based on this,he provides recommendation information to the user to meet the user's preference.The recommendation accuracy of the recommendation system is often Closely related to the recommendation algorithm.Therefore,if you want to improve the recommendation quality of the recommendation system,the best way is to optimize the recommendation algorithm.At present,most types of e-commerce platforms use collaborative filtering algorithms for recommendation,but as the amount of data increases and the level of information technology increases,the use of traditional collaborative filtering algorithms for recommendation is not sufficient to meet the demand.Constantly explore new and more efficient collaborative filtering algorithms.Based on the above considerations,this paper hopes to improve the traditional collaborative filtering algorithm and apply it to the recommendation system to improve the recommendation quality of the recommendation system.The creative work of this paper is to improve the data preprocessing stage of the collaborative filtering algorithm and to find the nearest neighbor collection stage.In the data preprocessing stage,one is to reduce the sparsity by filling the user-item scoring matrix;the second is to introduce the labeling factor and the time factor,so that the constructed user preference model has better expression effect.In the search for the nearest neighbor collection stage,the first is to combine the collaborative filtering algorithm with the binary K-means algorithm to select the user cluster matching the target user as the search range of the nearest neighbor set;the second is to improve the similarity measure formula.In the experimental part,the MovieLens dataset used in this article is a researchbased automatic collaborative filtering recommendation system developed by the University of Minnesota.The dataset contains 9000 users' 100,000 ratings of 1,682 movies,mainly used by users.The viewing record and scoring situation,recommend appropriate movies to different users' preferences.It is found through experiments that the improved collaborative filtering algorithm has lower error rate and improved accuracy than the user-based collaborative filtering algorithm,which indicates that the improved algorithm has higher recommendation quality.