Network Users Based On Web Log Clustering And Implementation

Clustering is an important mining method of data mining. It finds the comparability among the object from the database, and classifies the object to make different object in different kind as much as possible, and the object in the same class are as similar as possible, i.e." birds of a feather flock together ", thus optimize the implicit useful information or knowledge in the inquiry of the extensive database and discovery data, there is extensive application in a lot of fields in the data clustering.This paper mainly researches user clustering from Web Log.The results of user clustering can be used in optimizing the network structure and reconstructing the website and bringing the individuation and recommending.This paper analyses and researches user clustering which hides in Web Log from two aspects:the structure and content of user browsing paths,analyses the status of user clustering at present.Analysing UBPC algorithm is emphasis, pointing out the problems of applying this algorithm to user clustering and proposing improvement accordingly.This paper have finished user identification,conversation identification,session identification which are Log data pretreatments. In the process of user clustering based the structure of user browsing paths,it introduces a method of accounting the simular values between ordered non- numerical data.In the process of user clustering based the content of user browsing paths,it introduces the conception of object page and navigation page,proposes a new user clustering method with the result of object page clustering.This paper applys the improved user browsing path clustering algorithm to the processes of the user clustering above. Finally, on the analysis of the experiment results,it gives some advice about optimizing the network structure and the service of individuation.