Discovering User Interests Based On Bayesian Network

With the popularization of computer, the network becomes an essential part of people's life. Nowadays, the online shopping gives us a kind of convenient life style. Huge information from internet brings more choices to people, it causes some puzzles too. For customers how to select the most suitable thing from thousands of products, for enterprises how to improve customers' loyalty and keep the customer, how to meet different individualized demands of various kinds of users, etc., these questions become the most important things need to be solved. In order to make customers feel the totally individualized service and meets their demands, the more important thing is to understand users' interests, find customer's potential purchasing demand.Firstly, this paper introduces the personalized service which offers different services to different users to meet different demands. Personalized service study users' interests and behaviors by collecting and analyzing users' information, thus realize the purpose to recommend voluntarily. Secondly, the paper introduces Bayesian network and customer's loyalty. Bayesian Network offers an effective method to achieve personalization service. It combines knowledge and data, and find user's interest by directly probability relation. It also makes accurate prediction while interfering. The research of customer's loyalty is set up on the base of the customers' satisfied theory. The final purpose of strengthen customers' satisfaction is to improve customer's loyalty. So the research on loyalty can explain and reflect users' shopping behavior properly. The paper analyzes users' real-time behaviors, and turns the behaviors to users' initial interests, after this, it makes advanced research on users' initial interests by Bayesian Network. Because Bayesian Network can forecasts sudden events, it is more suitable to reflect users' interests. On this basis, this paper combines users' historical consumer behaviors, researches users' shopping habits by dynamic Bayesian Network and finds users' potential interests. In this way, not only the model considers the whole factors that influence users' interest, but also simplified the interest models. It meets the users' demands in the large degree.To satisfy the user's demands, based on the concept of customer's loyalty, the paper studies the structure of the network with huge data to update the interest model. Because the customer's loyalty has combined users' real-time behaviors and historical behaviors, and it reflects potential interests, the update model avoids the training of sample data, saves time for online recommendation.This paper combines historical behaviors with the real-time behaviors, it sets up a user's interest model effectively, and adds parameters for the structure of network to satisfy the changing interests and personality recommendation.Finally, it points out the deficiency, and looks forward to the direction that can be studied further in the future.