| In this paper, the developed situation of Web log mining and existing problems were systematically elaborated, according to Web log mining process and related technologies are introduced and discussed. Focus on analysis and research of the Web log data preprocessing, clustering analysis technology and algorithm such as Web log mining key technologies.This paper summed up systematically and studied the theory of fuzzy clustering, the basic principles of self-organizing neural network's algorithm. Respectively, discussed on issues such as clustering algorithmson, the fuzzy multi-matrix, self-organizing neural network training.At present, the study of Web user access pattern mining generally using a single algorithm, consider a single property when user accessed the page, allowing users accessed model results have a certain one-sidedness, unreasonableness. On this basis, in order to better improve the Web user access pattern mining accuracy, reduce the running time of user accessed patterns, this paper presents a Web user access patterns mining algorithm based on multiple fuzzy matrix_SOFM, under the comprehensive consideration of a variety of evaluation factors when user accessed Web to construct multiple fuzzy matrix, a better integration of several related SOFM neural network carried the learning and training on multiple evaluated factors, it is more accurate and scientific mine WEB user access patterns. at the same time, achieve dimension reduced operation on the fuzzy matrix through self-organizing neural network, accelerate the speed of the model, increase the anti-jamming, anti-noise function.The experimental results show that the proposed Web user access pattern mining model of multiple fuzzy matrix_SOFM, to a certain extent, improved the accuracy and operational efficiency of the WEB user access pattern mining, it is a good WEB user access pattern mining method.Finally, deficiency and achieved results have been summarized, and future research work is prospected. |
PDF Full Text Request