Reconstruction of user sessions from Web server log

Services provided online today are predominantly intelligent and user-centric. When a user uses a website, the site makes suggestions to user by studying their behavior. An important source for studying the user behavior is the access log file that records users´ requests. The behavior of customers is mined from the web log file present in the web server. The requests from the users are grouped into sessions based on heuristics. The common heuristics are time-oriented, navigation-oriented, link based or a combination of these. There are also approaches based on the integer programming model where sessionization is modelled as an optimization problem and uses techniques like Simulated Annealing as the heuristic for sessionization. This paper proposes a modified version of link based heuristic to handle back browser movements. It also makes use of the time heuristics. The proposed system is compared with an existing system that applies links heuristics for consecutive records and is found to be better.