Research On Application Of Web Usage Mining In Electronic Commerce Recommendation System

The rising popularity of electronic commerce makes data mining anindispensable technology for business competitiveness. Customers' access producesabundant raw data in the form of Web access log that is stored in Web server. Withoutdata mining technology, it is impossible to make any sense of such massive data. Inthis thesis, we focused on Web usage mining because it helps most appropriatelyunderstand users' behavioral patterns, which is the key to successful electroniccommerce recommendation system. Web Usage Mining is the application of datamining techniques to Web logs files in order to produce results used in some aspects,such as electronic commerce recommendation system.Firstly, a framework of electronic commerce recommendation system waspresented. Then its every module's function and how they correspond and worktogether was expatiated. Data preprocessing and frequent patterns mining werefocused. Data preprocess is a critical step in Web Usage Mining. The results of datapreprocessing are relevant to the next steps, such as transaction identification, pathanalysis, association rules mining, frequent patterns mining, and so forth. Analgorithm called USIA is presented and experimentally evaluated that its efficiency ishigh and it also can identify user and session exactly. A simple and efficient algorithm called Predictor was presented. It can mineassociation rules and frequent patterns effectively and correctly. It can satisfy the needof real time Web page recommendation and also can be used to incremental mining.Experiments conducted on real Web server logs verify the usefulness and practicalityof our proposed techniques.