Research On Collaborative Filtering Recommendation Algorithm Based On Similarity Calculation

Under the background of big data era,the huge amount of data from the Internet has caused the problem of "information overload",and the users cannot get the valuable information quickly,however the occurance of recommendation system alleviates this problem.As the core of recommendation system,collaborative filtering algorithm has gradually become the focus of experts and scholars at home and abroad.However,the traditional collaborative filtering algorithm has some problems,such as data the problems of sparsity of the data,cold start and scalability,which affect the quality of recommendation.In view of this problem,this paper studies and improves the traditional collaborative filtering algorithm,please check the main work as follows:(1)An improved algorithm based on the proportion of users' common scores and item attribute as the weight is proposed,aiming at the problem of inaccurate recommendation of traditional collaborative filtering algorithm,.While the traditional similarity algorithm measures the similarities between users based on the user common score item set,but cannot make a reasonable recommendation in many cases,so this paper puts forward a new way of thinking,considering the impact of users on the proportion of common scoring items,analyzes the interest of users expressed by the score size,combining the similarity based on item attribute preferences to form a new improved algorithm.Compared with other algorithms and traditional algorithms,the experimental results show that the improved algorithm has better performance,overcomes the shortcomings of the traditional algorithm,and improves the recommendation accuracy.(2)An improved collaborative filtering algorithm combining user activity is proposed with regards to the problem of data sparsity,The similarity measure between users is the key of the collaborative filtering algorithm based on users.The common similarity calculation methods will be greatly affected in sparse data sets,resulting in the accuracy of the recommended results have been reduced.In order to solve this problem,this paper proposes a new computing scheme by introducing user activity and considering the common score number and score difference of users,and optimizes the key performance of similarity measurement in collaborative overplay algorithm.Experimental results show that the improved algorithm can significantly reduce the average absolute error,thus effectively reduce the sparse data set on the recommendation accuracy,and significantly improve the recommendation quality of the recommendation system.