Hybrid Recommendation Algorithm Based on Improved Collaborative Filtering and Bipartite Network
Wenbin Jiao, Bingxian Ma, Lianjiang Zhu · 2020
Aiming at the problems of sparse data, low scalability and cold start in collaborative filtering and the problem of poor interpretability in bipartite network, this paper proposes a hybrid recommendation approach that combines collaborative filtering and bipartite network to capitalize on their respective strengths. First, we have optimized the sparseness, scalability, and similarity calculations of collaborative filtering through Weighted Slope One, K-means based on density, and punishing popular items. Then the bipartite network is optimized by weighting and nearest neighbor. Further, by defining an adjustment factor, we merge collaborative filtering and bipartite network to obtain a hybrid recommendation algorithm. Experiment shows the hybrid algorithm is feasible and can improve the accuracy of recommendation effectively.