Research on recommendation algorithm based on collaborative filtering of fusion model

Abstract

Based on the facts nowadays, recommendation systems are widely used, the focus of this research is how to effectively improve the accuracy and adaptability of such systems. The research proposes a collaborative filtering model algorithm based on the fusion model, which combines five classical algorithms: Singular value decomposition, Knn-Baseline, K-Means, Non-negativistic matrix factorization and SlopeOne algorithm. By combining the outputs of five models, and then carrying out regression and fusion, a collaborative filtering algorithm is obtained. The model can effectively integrate the advantages of the five models, and can carry out experimental validation on the datasets of Movielens-1M and Movielens-100k. The experimental results are more accurate than applying the five algorithms individually, and the regression model has good adaptability and predictability.