Intelligent Mining on Purchase Information and Recommendation System for E-Commerce

As an important marketing tool, recommendation systems for e-commerce offer an opportunity for merchants to discovery potential consumption tendency. This paper puts forward a novel recommendation algorithm to make the recommendation system more accurate, personalized and intelligent. Firstly, we use intelligent mining on purchase information, and regress consumer preference rating on click behavior. Secondly, we use Bipartite Network Recommendation model based on resource allocation and improved collaborative filtering model; the former abstracts products and consumers into nodes in the graph, and finds the correlation of products that recommend to others using alternative relation; and the latter solves the problem, caused by sparse data, by compressing rating matrix and predicting null values. Finally, according to Alibaba e-commerce customers purchase data, we verify that Hybrid Recommendation Model optimizes the accuracy and coverage of the recommendation results.