An Improved Collaborative Filtering Algorithm Based On User Interest Value And Expectation

In an environment with massive data,the personalized recommendation system becomes a weapon that helps users discover items or information of interest,and saves the users' time and effort in finding items or information to a large extent.On the other hand,with the rapid growth of data,the traditional collaborative filtering recommendation algorithm,which is one of the most successful recommendation algorithms,is facing the influence of increasingly severe sparsity problem of the user scoring matrix.In this paper,the recommendation accuracy of the recommendation system is analyzed and studied.Based on the User-Based Collaborative Filtering(UBCF)algorithm,the user similarity algorithm is improved by using the attribute of user interest,and the traditional K-Nearest Neighbor(KNN)algorithm is simply optimized.Combining the two,an improved collaborative filtering algorithm is proposed.This paper first describes the development of personalized recommendation strategies at home and abroad,and analyzes several common recommendation strategies,compares their advantages and disadvantages,and focuses on the user collaborative filtering algorithm.Then,we analyze the key-user similarity algorithm of cooperative filtering recommendation algorithm,and select the Pearson algorithm,which is the best effect of UBCF algorithm,to further study.We found some problems such as the interest of users is not fully utilized and the same user is not consistent with the scoring scale between different project categories.In this paper,we introduce the concept of user interest and expectation,and propose an improved user similarity calculation based on user interest value and expectation value.Next,by analyzing the traditional KNN algorithm,it is found that if the sparseness of the data is too large,only a few or no users will be allowed to target the project after the nearest neighbor user set is selected for the target user.This will have a negative impact on the recommended quality.Based on this point,this paper presents a simple improved KNN algorithm.Following,the improved user similarity algorithm and the improved KNN algorithm are combined to propose an improved collaborative filtering algorithm.And through the experiment to get the most appropriate harmonic parameters,based on this harmonic parameters,designed a series of reasonable experiments to verify the effectiveness of the algorithm.The experimental results show that the improved cooperative filtering algorithm proposed in this paper can improve the accuracy of user similarity to a large extent and get a better recommendation effect.Finally,based on the improved collaborative filtering algorithm,this paper designs and implements a simple prototype recommendation system,and completes the concrete application and implementation of the collaborative filtering algorithm.