Applications To The Analysis Of User Online Behavior Data Mining

With the rapid popularization and development of information technology, thenetwork has brought great convenience to people, but meanwhile, it produces largeamounts of data every day. However, in these huge amounts of data, vast majority of whichis junk data, only a very small part of the data has potential value for us. Therefore, how toeffectively extract the potentially hidden and valuable information from the massive useronline behavior record has become a hot topic of data mining research in the field, but alsoa problem has to be solved in today’s information age.As the network data is usually in various forms and its structure is complex, as well asthe amount of data is very large, the use of traditional relational databases such as SQLServer、MySQL, etc., is difficult to be accurately analyzed, meanwhile, the process ofanalysis is inefficiency. As the development of the relational database, data warehousetechnology commonly used to organize and store large amounts of heterogeneous networkdata. It is very intuitive and very easy to get the corresponding statistics for furtherexcavation and analysis for the data cube while analyzing the network data. What weshould do is just to extract data in the appropriate fields, build the mining analysis modeland multi-dimensional data cube from the data warehouse firstly. This paper, carried outunder the background of the actual development project, with the log of Web Servercollected from the Web named “Traces available in the Internet TrafficArchive--BU-WEB-Client dataset” as experimental data sets, throught using the OLAPonline analytical processing technology to mine and analyze the dataset stored in the datawarehouse from different dimensions and particle size. And finally, the user onlinebehavior characteristics for the network users such as the user’s online period and time aswell as the access file type, etc., is succeed to dig out. Meanwhile, setting the onlinebehavior of accessing the Internet movie sites as an example, using the time seriesprediction algorithms of ARMA (p, q) to establish a forecasting model, and the number ofpeople accessing movies sites within the next two weeks were predicted. Experimental results show that, in the case of small step, the model has a good prediction to the numberof Internet users in a relatively short period of future, with a high degree of fit between thepredicted value and the actual observations, and it is a reliable prediction model to predictthe number of users accessing movies sites, which provide the decision support andscientific basis on the network management, maintenance and optimization for the LANmanagement.