Research On Recommendation Algorithm With Diversity Based On Two-stage Neighborhood Selection

Recommendation system plays an important role in the field of e-commerce.It can help users screen out the content that users are interested in from the huge database and recommend it to target users.Collaborative filtering-based recommender systems are important tools for predicting user preferences by analyzing their historical rating information.However,traditional collaborative filtering methods primarily exploit onestage neighborhood selection to provide accurate recommendations,while often neglect the diverse demands for recommendations of users.How to improve the diversity of recommendations while maintaining accuracy is an important research problem of recommendation system.To address the tradeoff between accuracy and diversity,a novel method based on two-stage neighborhood selection is proposed.This method considers both the similarity between users and dissimilarity between neighbors in the neighborhood selection phase.In the first stage,the similarity between users is calculated according to the Jaccard index,and the users most similar to the target user are selected as neighbors.In the second stage,an improved product-moment correlation coefficient is used to obtain the attribute preferences of a user on the rated items;then,the scope and ranking of the user preferences are determined.And,a preference dissimilarity model is constructed to evaluate the preference difference between users and obtain a preference dissimilarity matrix according to the scope and ranking of the preferences.The preference dissimilarity submatrix corresponding to the neighbors selected in the first stage is extracted,and the hierarchical clustering of neighbors is carried out according to the submatrix,and the diversified neighbors are selected based on this.we adopt a two-stage neighborhood selection method to identify a set of neighbors that are dissimilar within themselves and similar to the target user.The aim of improving diversity is achieved on the premise of ensuring accuracy.Experiments are carried out on two datasets from different sources,and the experimental results show that the proposed algorithm performs better than the comparison algorithms considering accuracy and diversity.Thus,our scheme provides a new insight for improving the accuracy and diversity of personalized recommendations.Finally,the applicability of the proposed algorithm in information procurement scenarios is analyzed,and the practical significance of the algorithm is verified.