To help consumers find the most wanted products effectively, e-commerce recommendation saves lots of time spent on viewing unnecessary web pages and increases revenue for ecommerce websites. Because of this significance, this paper concentrates on technology of recommendation and makes an improvement to recommender system. To solve the problem of current technology that only concerns association between two products, this model considers the product network and guides consumers to view products following the intended path. The object is to maximize revenue of the entire product network. An empirical study of yhd.com shows our model is more effective than current model.