Research On Personalized Recommendation Service Based On Clustering Of Web Access Log

With the rapid development of Internet technology, the network has accumulated a wealth of information, resources and all kinds of applications, and is still used in the rapid growth. This brings two questions:for the web site managers, they should consider how to use these resources to find out the hidden knowledge, provide users with better services and improve the competitiveness of their products. As the same time, the web users cannot find the knowledge they need when they face with a deluge of information resources. They hope that the site can be able to push the interest information to them, resolving the problem of "information overload" and "lost in information".In recent years, foreign scholars have used web data mining to solve these problems. they have made a lot of effective papers about data mining techniques used in internet products and applications. We can use the vast amounts of user behavior logs and site access logs to find the users' interested in web resources. We can also use the logs to forecast the user's access trends. As a research field, Web data mining has become very active in recent years.In this paper, we use the web users' access logs as a data source, based on the theory of data mining and personalized recommendations, proposed a personalized recommendation model based on K-Means clustering algorithm:First, use the improved K-Means clustering algorithm to divide the user into different clusters by interest on websites, and then through the Collaborative Filtering recommendation to find most interested sites to recommend in a usergroup. Through this two process, we can not only ensure the accuracy of recommendations, but also improve the efficiency of personalized recommendation; On the basis of these studies, I also try to make the design realization, analyzed the experimental results, and follow-up Application direction and improvement points.