Web log mining has become one of the most important applications of data mining. Mining web logs can reveal important information of how the web is being accessed, and can provide information for many tasks ranging from improving web appearance to categorizing potential buyers. There are many techniques available to mine information in web logs. This paper presents a unique approach to cluster and classify user sessions using a naive Bayes' classifier. The categories are generated using a modification of the classical fuzzy c means algorithm, SISC. Subtractive clustering was used as a predecessor to find the number of categories dynamically. Some experiments were reported and our algorithm was compared with hard clustering algorithms like k-means, which show the efficiency of our approach.
展开▼
