Personal Recommendation Based on Community Partition of Bipartite Network

In recent years, the recommendation system based on bipartite network has attracted much attention. However, there still exist some problems such as low precision and diversity deficiency in traditional algorithms such as network based inference (NBI) or probability spreading (ProbS). As there exists hidden information of community structure in bipartite network, we can take full advantage of the community partition results to improve the recommendation performance. In this paper, we introduce a new bipartite network partition way based on heuristic modularity optimization. We can get the User and Object community partition when the bipartite network modularity get to the maximum. Then we work out the recommendation list for each user based on the community partition results through four compound ways. Experimental results show that our algorithm achieves better performance both at precision and diversity compared with traditional recommendation algorithms.