Mining Interests and Navigation Patterns in Personalization on Portal

Personalization services pose new challenges to interest mining on portal, such as many powerful functions to customize desktops. Capturing these surfing behaviors of users implicitly and mining interest navigation patterns are the top demanding tasks. Based on the summary of personalized interest mapping on portal, a novel portal-indepeodent mechanism of interest elicitation with privacy preserving is proposed, which implements both the implicit extraction of diverse access behaviors and corresponding semantic analysis. A new schema is also presented to extend the interest representation rule efficiently in privacy preserving process. Then the legal transparent accountable interest in terms of the interest behavior entries could be implied with this interest-extended rule. Moreover, the navigation relationship among the interests can describe the user's next possible interest trends, especially benefiting the recommendation. By combining the association effect of those interests and the prediction on interest intentions, a hidden Markov model is extended with personalized interest descriptions of portal to form interest navigation patterns for different users. Then experiments are carried out in order to validate the proposed approaches with availability and feasibility. The improvement of representation accuracy and mining capability for the complex interests on portal is a feature that clearly distinguishes the proposed approaches from traditional ones.