Clustering-Based Analysis of User Behavior Logs in Web-Based Software Applications

Abstract

User behavior logs from web-based software applications provide detailed evidence of how users navigate pages, submit forms, trigger errors, search for information, and complete operational tasks. This article presents a clusteringbased framework for analyzing web application behavior logs using sessionlevel features derived from clickstream and event data. The framework cleans raw logs, reconstructs user sessions, extracts behavioral indicators such as session duration, page depth, interaction diversity, form submissions, error events, repeated navigation, and exit patterns, and groups similar sessions into interpretable user behavior clusters. The study shows that clustering can separate efficient task completers, exploratory navigators, transaction-heavy users, and friction-prone users without requiring predefined user labels. The proposed approach supports application improvement by identifying where users complete tasks smoothly, where they explore heavily, where workload intensity is high, and where repeated errors or abandonment patterns indicate usability or workflow problems.