Recommended System Based On Improved Forecasting Markov Model

With the rapid development of Internet technology, network has penetrated into the every aspects of our lives. At the same time, as the Internet information explosed increasing, more and more website operators began to paying attention to the personalized information services of websites to make users of the websites quickly find the information they want from the huge amount information, meet users’ demand for browsing efficiency.Predictive Recommendation System emerged in this background. Predictive recommendation system is a system which uses data mining, artificial intelligence technology to provide intelligence and personalized services for users according to user’ browsing histories and the website’s need. So that it can have more accurate rate of users’recommend pages.The core of this system is the method of prediction which is the focus of this paper. After reading a lot of literature about prediction methods, we found that Markov prediction model is simple and easy, and suitable as prediction model for intelligent recommendation system. At the same time we also found that Markov model has a contradictions between accuracy and computation overhead in practical application. In order to solute this problem,this paper propsed a method which mixed Markov, clustering method and ant colony optimization, then used the time that users spent on browsing and the times users clicked on one page as the prediction reference. The simulation results show that the new method can ease the contradiction mentioned above. At last this paper also proposed a new quantitative criteria to evalue the function of the website after using predictive recommendation system based on queuing model.