Cluster Research Of Web Log Mining Based On Rough Set And Self-organized Neural Network

Web Log Mining is defined as retrieve browse model from Web server-side logs or users' browse record. Web Log Mining could help to get users' browse model which could be used to improve the website structure or provide users with personalized services.In this paper, cluster analysis was used to retrieve information from Web Log, and then provide personalized service to users by the principle that the people in the same cluster have the same interests. Cluster analysis was completed by Self-organized Neural Network and Rough Set in this paper. Self-organized Neural Network is a kind of Neural Network which was organized in some shape recognizes the model by cooperation and competition between the neighboring neurons. Self-organized Neural Network is learning by the competitive learning strategy which chooses the only winning neuron, but in fact there is a group of wining neurons.Rough Set which has lower approximation and an upper approximation could resolve this disadvantage. The self-organized neural network based on Rough Set could resolve the wining neuron uncertainty. First, select the best winning neuron, and then choose the second winning neuron set by some strategy. If the second winning neuron set is not empty, the best winning neuron and the neurons in the second winning set are the winning neurons, different learning strategy will be used point to the different winning result. Experiment prove that, compare with the Self-organized neural network, the Self-organized neural network bases on Rough Set cluster results are more averaged and have little dead neurons.Trained neural network will be used in pattern recognition. The Self-organized neural network based on Rough Set could recognize more than one interest from the users. It is prove that the Self-organized neural network based on Rough Set perform better both in learning and pattern recognition.