Research On Prediction Of Users' Navigation Pattern Based On Markov Model

Web log mining needs to summarize and predict the users' navigation pattern. And the Markov model is a simple and practical tool to do that. But some existing prediction methods based on Markov model still have some shortcoming. So it becomes a new lesson in the area of web log mining that how to improve predicton methods. This paper analyses the current domestic and international research results of how to use Markov model to predict the users' navigation pattern. Then we find some problems of existing prediction methods based on Markov models on work area and cost of time and present our plan to correct them. And we study the improving of prediction methods based on Markov model.First of all, this paper introduces the principle and work process of the traditional Markov model. By analyzing the input and output data of traditional Markov model, we find that it can not reflect users' interest exactly sometimes. And then with the way of aggregation of multi-dimensional hierarchical data, a prediction approach for webpage types is presented in this paper. Finally, the efficiency of the approach is validated through the experiment.Second of all, this paper studies multi-Markov chains model which use cluster to make the users different types. Then we analyze its algorism, and find that multi-Markov chains model can give higher prediction accuracy and requires lower space complexity than traditional model does, but it costs too much on time complexity. For that, this paper presents the dynamic-Markov model and gives out three algorisms for dynamic-Markov model. Then we prove that our algorisms are much better at time complexity than the one of multi-Markov chains model.Finally, we analyze the traditional Markov model, multi-Markov chains model and dynamic-Markov model through experiments. And we present the space complexity of dynamic-Markov model based on experiments. The results of our experiments validate efficiency of the new approach.