Forecasting website accesses using data mining techniques

Abstract

In the current business context, evaluation of the web page views or web pages accesses is gaining a huge recognition where it has become a vital need in planning of computer resource allocations and estimation process of upcoming revenue and advertising growth. One of the essential tasks of the hosting service provider is to allocate servers to each of the websites to maintain a certain level of quality of service for different levels of incoming requests at each point of time, and optimize the use of server resources, while maximizing its profits. In order to have a proactive management of resources within the servers it is required to build an accurate forecasting process of web accesses per unit time. As a time series, the web access patterns of the users exhibit not even different levels of short time random fluctuations but also periodic patterns that evolve randomly from one period to another. In this paper, we focus on extracting trends and web access patterns from page view or page access series using data mining techniques and their applicability on predicting the future web page accesses. Based on the time series regression techniques which are used to analyse and forecast web accesses, we found that Sequential Minimal Optimization and Linear Regression algorithms were providing more accurate results in forecasting.