Collaborative Filtering Recommendation Algorithm with Item Label Features

Abstract

The traditional item-based recommendation algorithm only considers the user's score when calculating, but the actual user's score has a malicious evaluation, which seriously affects the accuracy of using the score for prediction. At the same time, the user's evaluation is small, and the amount of usage will also affect the accuracy of the recommendation. Aiming at the above two problems, this paper proposes a collaborative filtering recommendation algorithm with the characteristics of the item label. Firstly, the item label feature and the user's behavior data are comprehensively considered. Secondly, the user's feature data on the label selection is calculated, and finally the item is calculated in combination with the feature data. Similarity, thus making recommendations for users. Experiments show that the algorithm can solve the cold start problem of data well, and the interpretation of the recommended results is also convincing.