Logging Web Behaviour for Association Rule Mining
Ilkka Lilleberg · Theseus (Ammattikorkeakoulujen) · 2015
The aim of this study was to suggest improvements to web server logs of existing web services. The ultimate goal was to study how logging process could be developed to enhance using log data with data mining tools to get better results. Better results could be, for example, understanding which actions during web browsing session indicate interest to buy certain products in an online store. Data sets studied consist of web server data from Nasa web service, Sonera Joulukampanja webserver data and web server data from liandersson.fi web site. Liandersson.fi and Sonera Joulukampanja are based on Linux, Apache or Nginx, PHP and MySQL technologies. For the study a data mining environment using association rule analysis to mine new information from web server log data was developed. The emphasis was on doing research on what kind of data mining results can be achieved with different kinds of log data and how developing the logging of web server data can affect the possibilities of data mining. The two stated goals of the study were to generate a tool set that can automate the analysis of chosen log files and produce association rules of the log data and to analyse the capabilities of this data mining process, with the data available and to create a list of actions points that can be used to develop data logging. Yet another objective was to develop the logging of web server data and thus improve analysing this data in the future. Alongside developing a data mining environment, the key outcome of the study is instructions of how to plan web server logging in a way that takes into account the requirements of data mining of the specific data in question in order to use the information created by data mining to improve existing web services.