Merging Attribute Characteristics in Collaborative Filtering to Alleviate Data Sparsity and Cold Start

Abstract

Collaborative filtering algorithm is one of the most important technologies in the current e-commerce recommendation system. The effect of the similarity measurement method directly determines the accuracy of the recommendation system. Aiming at the shortcomings of the similarity calculation of traditional Item-based collaborative filtering recommendation algorithm in the case of extremely sparse user score data, an Item-based collaborative filtering algorithm based on attribute similarity is proposed. The algorithm uses the similarity of project attributes to correct the original similarity calculation, comprehensively considers the impact of project attribute values and user evaluation on recommendations, and dynamically adjusts the calculated values of similarity according to the sparseness of the score data to truly reflect each other similarity. The experimental results show that the improved algorithm can significantly improve the recommendation quality of the recommendation system when the user score data is extremely sparse, and can solve the recommendation problem of the new project.