User Behavior Analysis Based On Web Log

With the rapid development of computer technology in recent years,that the WEBapplication is alse develop so fast.so the web-user is more and more important,in theother way,the web-user is increasingly high requirements on WEB,such as:WEB need tohave intelligent, fast, and accurate to help users find the information theyneed;Meanwhile, as the operator of the WEB SITE, user behavior on the WEB areincreasingly interested. After understanding the web-user’s behavior,that can provide amore accurate service,such as:Precise Marketing.Also, according to the web-user’sbehavior, we can organizational and structural adjustment the WEB,in order to attractmore and more users.Furthermore, it can be based on WEB user’s behavior, to analyzethe deficiencies on the WEB SITE,for the safety and to provide a reference andfoundation.This thesis is based on the "ceramics e-commerce for the researchplatform",as it is as a sub-platform which named "Ceramic cluster area informationservice platform",which is a national863project.And this platform is main toachievement of the online e-commerce functionality.In this thesis,the research is based on the WEB Logfile,by using the regularexpressions to analysis it. Algorithms are mainly used Apriori algorithm,that to calculatethe association rules and K-means clustering algorithm is used to calculate the users.Meanwhile, the two algorithms are also some improved algorithms are analyzed andcompared, such as improved Apriori algorithm preference path algorithm, K-means ink-means++algorithm, based on the weight of the k-means algorithm.After this,it canget the access of the user’s affairs.Then using the knowledge association rules,classification, clustering and other data mining method to analysis the user’s transactionassociation and clustering.This thesis aims to analyze the user information on theceramics e-commerce platform, to complete precision marketing,system optimizationand anomaly detection,provide the theoretical basis and data. Finally,by using thealgorithms over the application system,it gets some conclusions: The result according tothe merely strong rules is not enough of the Apriori algorithm,it need to calculate theraise,or the results obtained may be inaccurate. If the center pointer of the k-meansalgorithm is inappropriate,the result of the gap will be relatively large,or the result will bethe same with the k-means++algorithm.