Application Of Artificial Immune Clustering In Web Log Mining

The web log mining is to access logs from users recorded in the server-side or extract useful modes from information browsed by users, and those data can help understand users' behavior and finally lead enterprises to wise market decision, providing better personalized services for users. In this thesis, the artificial immune ideas is adopted to do clustering, extracting users access mode from web log files and provide personalized services for them.Log mining technology and its procedures are studied practically in this thesis by selecting logs from the school of humanities in Nanchang University as the data source and the Weka as the platform. Research on the log mining technology and its important algorithms, introduce the process of data pre-processing detailedly, and experiment with them to achieve at the same time.Facing the deficiencies in traditional clustering algorithms nowadays, the analysis to the web log clustering is discussed adopting the artificial immune ideas, based on studying clustering algorithms, and the characteristics of immune system and the features of web logs data. Using the self-organization and self-adaptability of the immune system itself, and the combination of the principle of clone selection and that of the immune network, the clustering is realized in this algorithm in the context of antibody collection, by that category tags are added to every antibody in order to let it gain the full characteristics of the antigen, forming the memory antibody which can represent that antigen. The feasibility and accuracy are also proved by the verification and comparison in experiments, which sheds light on it theoretical and practical meaning.At last, through the result of Clustering analysis, provide the suggestion of modifying the structure to Web site managers, and then they can provide users with better personalized service.