Personalized Recommendation System Based On Web Usage Mining

Information overload and information disorientation are obstructions which restrain people from using internet information efficiently. Information overload means that it is difficult for people to absorb and understand information when facing so much data. Information disorientation refers to the state that people puzzle how to express and search information that they need. Personalized recommendation system can recommend information automatically according to users' interest. It solves the knottiness that people trouble to find information.The prototype of personalized recommendation system based on web usage mining is designed and realized. It mines users' interesting browse pattern from web log and offers recommendation service online to guide people browsing. Three modules are given. 1) Preprocessing of the web log. Preprocessing which includes data cleaning, user recognition, session identification, path supplementation and user transaction pattern recognition is the first stage. The paper discusses preprocessing in detail, and gives a key algorithm of each step. 2) Mode mining. FCC transaction clustering algorithm and browsing paths mining algorithm are relized in this part. Confidence and preference are discussed and browsing paths mining algorithm is ameliorated. 3) Online recommendation service. Comparing user's accessing sequence and mining results offered by mode miming part, the system provides recommendation pages through recommendation algorithm. Including accessing order and association rule, recommendation algorithm based on transaction clustering is improved.Last, the paper summarizes the deficiency of system and points out the orientation and perspective of personalized recommendation system.