| In the rapid development of information technology today,the explosive growth of data often leads to information overload and loss,how to mine valuable information from these massive data has become a hot spot.On the one hand,users can search for the information they need through categorized directories and search engines;on the other hand,recommendation algorithms are used to analyze user information,obtain information that users like,and then recommend it to users.Memory-based collaborative filtering is the main research method for recommendation systems due to its easy implementation,cross-domain compatibility and interpretability,but it has limitations related to cold start,low accuracy,and data sparsity,among which low accuracy and data sparsity are the most common problems.To address the above problems,the paper makes the following improvements:(1)To address the problems of current collaborative filtering algorithms in terms of low accuracy and data sparsity,this paper proposes a collaborative filtering algorithm that uses a fusion of improved similarity and high-density-first clustering algorithms.The algorithm firstly calculates the density of data points based on their distances and selects the point with the largest density as the initial clustering center to obtain K clusters in turn;secondly,it introduces the user’s rating time factor of items in the traditional similarity calculation to enhance the collaborative recommendation weight of recently visited items by users;finally,the algorithm increases the rating difference between users on common rating items and reflects the user’s interest preference in the calculation of user similarity.The experimental results show that the algorithm can effectively improve the accuracy and data sparsity problems of collaborative filtering algorithm,and improve the recommendation performance of the algorithm.(2)In response to the problem that the collaborative filtering algorithm ignores the trust relationship between users when calculating user similarity,which leads to low prediction accuracy,a method is proposed to deeply explore the role of trust relationship between users on user similarity and improve the shortcomings of traditional user trust degree calculation.The method introduces the response factor between users and calculates the direct trust degree based on the results of successful or failed interactions;meanwhile,considering the sparse data set and the possible existence of no common rating items between users,the trust transfer is used to obtain the indirect trust degree of two users.Finally,the improved user trust degree and user rating similarity are combined with the screening effect of the target user’s nearest neighbor set to further improve the accuracy of the recommendation algorithm.This article proposes a new algorithm to address the issues of accuracy and diversity in recommendation algorithms,and conducts experiments and comparative analysis on the Movie Lens dataset.Experimental results show that the proposed algorithm can significantly improve the accuracy and diversity of recommendation algorithms,providing a new approach to improve the effectiveness of recommendation algorithms and enhance user satisfaction. |
