Web Information Recommendation Based on User Behaviors

In our study, we propose a Web information recommendation framework depending on user behaviors. It is different from traditional recommendation systems depending on single association rule or classification engine. It has an interactive interface with people who could adjust his access manner according to the requirements by himself. It differs from the system which needs each reader to indicate whether or not he likes report in order to extract knowledge by offline analysis. In our system, people could select manners of recommendation, for example clustered information or associated information extracted from user activity history. Simultaneously, meaning clustered and associated information knowledge could be extracted accurately by recursion invoking between K-Means and Apriori. Therefore, surfers could initiatively discover various patterns meeting their requirements accurately.