The Investigation That Clustering Pretreatment&BP Neural Network For Classifying The Web Users

All aspects of the Web are developing rapidly. The Web is mainly used in sharinginformation in the early days, but now Web applications have spread to e-commerce, onlinegames, remote login and other fields. Also there is a wide range of applications in these fields. Ahuge of people use the Internet to query the data and information that they need. Although theInternet is developing so fast and providing people with increasingly large information resources,face to such huge of information to rely on manual search for the information, the efficiency isvery low and sometimes the result is inaccurate. Concern for their personalized become one ofthe modern network technology issues, increasing in earnings for people who design Web sitesuse their personalized arrangements advertising. Thus it needs a higher demand on the Web sitedesign and functionality. Nowadays the use of information about the Web site to cluster withsimilar access interest in the transaction set finishing and mining, efficient access to the user inthe access mode for the user to provide quality services or personalized has become the focus ofattention of business and academia. In order to solve these problems, some people study of theseproblems, and propose some the Web user clustering methods and the use of clustering as apreliminary empirical data for the artificial neural network, to improve the efficiency of the userclassification, and thus provide users with better service.It can mine the Web log with method of Web user clustering at first, this process is dividedinto three steps: firstly, log preprocessing is used to extract the user some basic features;secondly, followed by the extraction characteristics of the user to calculate the similaritybetween the individual users; finally, the most critical is user clustering. Then get the userclustering preliminary data for the transformation of the artificial neural network, to improve theefficiency and accuracy of artificial neural network. It is very important to extract the userfeature and compute the user similarity in the clustering process, because of their direct impacton the effect of user clustering, the effect of user clustering directly affect the accuracy ofartificial neural network. Currently, users of the feature extraction are based on a user session path, but the relationship between user interest and a lot of behavioral factors is too complex, it isdifficult to establish a mathematical model; the general similarity calculation between the sets ofordinary intersection operator, or some similar method, but these methods have computationaloverhead to calculate a long time, inefficient, and matrix huge there is no good solution; generalcluster computing is in this point, no further use of such information and data for deeper mining.For a series of questions above, the text sums the user navigation path up as a regressionproblem, this model can be directly quantified the users’ interest in Web pages, you can use someuseful compression algorithms to compress the matrix huge, and on this basis can be furtherefficient using the clustering algorithm. Finally, carrying on the efficient clustering algorithm.Used to complete the clustering of Web users based on the use of regression equation to quantify.And comparison experiments found that the algorithm can improve the accuracy of userclustering, the results verify that this algorithm is effective. Efficient Clustering use thisinformation to establish a self-learning artificial neural network.This is the improved clustering algorithm based on information to establish a suitableartificial neural network characteristics, and efficient to be more efficient clustering algorithmand a more powerful machine learning algorithms.