An Efficient Technique for Mining Usage Profiles Using Relational Fuzzy Subtractive Clustering

We propose an efficient technique for mining web usage profiles based on subtractive clustering that scales to large datasets. Unlike earlier clustering based techniques for the same purpose, our technique does not require user specification of any input parameter to obtain the desired clustering. Instead, we achieve this by searching in the cluster space for the best clustering of the given web usage data. To evaluate clustering quality, we have formulated a validity index for our algorithm. Our implementation of the proposed technique and the experiments with large real life datasets show that it indeed mines the desired usage profiles much faster than existing techniques.