Abstract
This paper presents the application of data mining algorithms to the prediction of Web performance. Our domain-driven data mining uses historic HTTP transactions data reflecting Web performance as perceived by the end-users located in the Internet domain of Wroclaw University of Technology, Wroclaw, Poland. The predictive modeling features of two general data mining systems, Microsoft SQL Server and IBM Intelligent Miner, are compared. The neural networks, decision tree, time series, and transform regression models are evaluated. It is shown that the data mining algorithms return quite accurate prediction results. The best results are achieved using the IBM's transform regression algorithm.
This work was supported by the Polish Ministry of Science and Higher Educaiton under Grant No. N516 032 31/3359 (2006–2009).