Research To The Pattern Mining Of Network Accessing Behavior And Its Application

With the informatization construction of petroleum industry, exploration information, oilfield development information, ground construction information, storage and transportationsales information, management information, and other petroleum information have formed acertain scale of huge amounts of data resources, and most of which are provided to users as aWeb site. How to better organize these information, to provide users with intelligent webinformation classification, and to meet the user's individual information needs, these problemsneed to be solved. In response to these problems, this text has presented the method ofpetroleum information patterns mining of user accessing behavior, by analyzing user accessbehavior to petroleum information, this method reveals the habits and interests hidden in logsof user access to petroleum information,and helps the user for convenient access to theinformation needed from a large number of petroleum information.The user accessing behavior patterns mining of petroleum information is mainly dividedinto three phases: data pre-processing, pattern discovery, analysis and application of pattern.In the data pre-processing phase, this text has proposed a improved method in which the sitehome was combined with dynamic session identification time threshold, this method canbetter extract the user's browsing path, and improve the accuracy of data pre-processingresults with the experimental verification. In pattern discovery phase, this text has adopted thek-means clustering algorithm, because of the shortcomings of the k-means algorithm stronglydependence on the initial cluster centers selected, the initial cluster centers were determinedwith the density and distance-based methods to prevent the clustering results from the impactof the initial cluster centers selected randomly. The improved algorithm has betterperformance in clustering accuracy, efficiency and stability with experimental validation.The user access information of a petroleum information resource network was adopted asbusiness data, and the cluster of the users of this site was completed, through the analysis ofthe clustering results, some useful patterns were obtained for providing the basis on petroleuminformation retrieve and personalized service.