Combining DBSCAN and MLP for Knowledge Extraction in Web Usage Mining

Abstract

The web has become an invaluable information source due to its vast volume and complexity, sparking the development of Web Mining methods to extract relevant knowledge and meet user needs. Among these, Web Usage Mining (WUM) focuses on analyzing user activity data to personalize content, improve site design, and enhance navigation. WUM involves key steps such as data preprocessing and knowledge extraction, applied to log files tracking user interactions. In this study, we propose a hybrid approach combining DBSCAN clustering for grouping user behavior and MLP classification to identify user patterns. By preprocessing log data and applying these techniques, we identified significant navigation patterns that unveil user preferences and behaviors. The results demonstrate how hybrid methods can optimize WUM by offering precise and meaningful insights, bridging the gap between complex web interactions and actionable outcomes. This work highlights the potential of integrating clustering and classification techniques to enhance decision-making and user satisfaction in digital environments.