Experimental Investigation of PSO Based Web User Session Clustering

Web user session clustering is very important in web usage mining for web personalization. This paper proposes a Particle Swarm Optimization (PSO) based sequence clustering approach and presents an experimentally investigation of the PSO based sequence clustering methods, which use three original PSO variants and their corresponding variants of a hybrid PSO with real value mutation. The investigation was conducted in 45 test cases using five web user session datasets extracted from a real world web site. The experimental results of these methods are compared with the results obtained from the traditional k-means clustering method. Some interesting observations have been made. In the most of test cases under consideration, the PSO and PSO-RVM methods have better performance than the k-means method. Furthermore, the PSO-RVM methods show better performance than the corresponding PSO methods in the cases in which the similarity measure function is more complex.