A Collaborative Recommendation Algorithm Based On User Similarity Improvement And Neighbor User Search Strategy Optimization

With the rapid development of Internet technology and information industry,the information overload has become a severe problem.A very effective way to solve the problem of information overload is a personalized recommendation system,which can recommend information or products to users according to their information needs or interests.However,the traditional collaborative recommendation algorithm has some problems such as sparse data,real-time performance,accuracy,and cold start.In this paper,we make deeply analysis and research on the two key steps in the collaborative filtering algorithm:user similarity computation and the search of neighbor users.Two optimization schemes are proposed to solve the problems of user rating scale differences and the taditional KNN neighbor search algorithm.Firstly,on the research of the user-item rating matrix,we found that different users have different rating habits,which led to the existence of rating scale differences between users.In this paper,we propose a weighting factor based on rating scale difference problem to improve the accuracy of user's similarity.Secondly,because of the data sparsity of the user-item rating matrix,when the traditional KNN algorithm is used to search the neighbor,there are two problems,one is that the similarity with the target user is high,but the selected neighbor hasn't rated the target item,and the other is that the selected neighbor has rated the target item,but the similarity with the target user is low,not the target user real neighbors.In this paper,the similarity threshold and weighting factor are used to combine the two strategies to calculate the final prediction rating,which can improve the recommendation precision.Finally,several experiments are designed on the Movielens dataset,which are compared with the traditional recommendation algorithm.The experimental results show that the average absolute error MAE of the improved similarity algorithm and the nearest neighbor search algorithm is superior to the traditional recommendation algorithm.So the results show that the improved recommendation algorithm has higher recommendation accuracy than the traditional collaborative filtering recommendation algorithm.