Personal Recommendation using Weighted Bipartite Graph Projection

This work is a study of personal recommendation algorithm employing the projection of weighted bipartite consumer-product network. The weight of the edges is directly the rate that a customer giving on a product. Following a network based resource allocation process we get similarities between every pair of consumers, which is then used to produce prediction and recommendation. We show this is also a two step random walk process in the bipartite. Since the weighted graph is more informative, we would expect higher predict accuracy.