Collaborative filtering recommendation algorithm based on both user and item

Abstract

Based on the analysis of the sparse problem and the cold start problem in the traditional collaborative filtering recommendation, a new collaborative filtering recommendation algorithm based on adaptive nearest neighbor selection is proposed. The algorithm considers the influence factors of user characteristics and item attributes, and then calculates the nearest neighbor sets of target users and target projects by using the score similarity model. According to the situation of the sparse score data, the similarity measurement results of two aspects are handled by the adaptive coordination factors, so as to get the final project forecast score. Experiments show that the proposed algorithm can effectively balance the instability effects based on the user group score and the recommendation based on the item group, and effectively alleviate the problems caused by the data sparsity. The experimental results show our method can increase the data density and achieve lower MAE (Mean Absolute Error), that is to say, the proposed approach can efficiently improve recommendation quality.