Research On Slow Selling Products Recommendation Techniques In E-Commerce System

With the rapid development of E-commerce and Internet technology, the number of products increase sharply. In this context, recommendation systems and personalized recommendation technology have emerged. Recommendation system can help customers to search their interesting products. And it plays a role of holding old customers, attracting new customers, and improving customer satisfaction, so that increasing the quantities of sale. But with more and more products emerge, some became slow selling goods due to the users finding them hardly, but they still have the value of themselves. This phenomenon will result to material waste and loss of profits. Mining the values of slow-selling products can not only reduce economic loss, but also have greater economic benefits for enterprise due to low cost. Therefore, the recommendation technique of slow selling products in E-commerce plays an important role in E-commerce system.This thesis firstly introduces some personalized recommendation techniques. Collaborative filtering and association rules are mainly described.Secondly, a recommendation model of slow selling products in terms of intermediate product is proposed. Intermediate product is in connection with the unmarketable goods closely, which is used to encourage purchase of the unmarketable goods and is mined and predicted by analysis of historical sales data. A method of personalized recommendation is proposed according to intermediate products interestingness with time weight. It can predict the slow selling products effectively and provide personalized recommendation according to customer's character.In addition, a recommendation algorithm based on profit maximization for slow selling products is presented. The main idea is to select the maximal profit item based on association between products, then recommend the items of maximum profits to every custom. And each item consists of one intermediate product and one slow selling product.Finally, experimental results prove the validity of the slow selling products predict and recommendation algorithm based on profit maximization for slow selling products.