Productrank: A Random Walk Model for E-Commerce Recommendations

Electronic Commerce has offered a convenient way for people to go shopping on the Internet. However, it is difficult for Internet customers to select a valuable item from the great number of various products available on line. When we use a keyword and search in a EC website, the ranking algorithm of products is usually based on statistics or simply the shop manager's preference, which does not fully exploit the knowledge and experiences hidden in the prior daily transaction records in the database. In this paper, we propose a novel approach to extract the purchasing behaviour of customers who purchase the same kind of goods, and with which we rank the products for user personally by comparing their behaviour. We introduce our evaluation metrics to assess the prediction accuracy of the proposed recommendation algorithm using transaction records of an online wine shop, the experiment results show that our algorithm is able to produce valuable recommendations.