High-Accuracy Predictive Modeling and Anomaly Detection in Web Portals

Article & Journal: Predictive modeling and anomaly detection in large-scale web portals through the CAWAL framework Published in: Knowledge-Based Systems

Study Summary: Building on the CAWAL framework, this study explores its potential in AI-driven applications. We utilized the enriched datasets generated by CAWAL to train machine learning models for two critical tasks: predicting the next page a user will visit and detecting anomalous (bot or attack) behavior. The results showed that models trained with CAWAL data achieved significantly higher accuracy (over 92%) compared to models trained with standard server logs.

Behind the Research: Data quality is the fuel of artificial intelligence. Standard logs are often too noisy for precise training. Because CAWAL captures "verified" human behavior and session details, the resulting datasets are incredibly rich features for machine learning. We demonstrated that this framework isn't just for reporting; it is a powerful engine for real-time security and predictive user assistance. It acts as an early warning system against sophisticated application-layer attacks.

Citation & DOI: Canay, O., & Kocabıçak, Ü. (2024). Predictive modeling and anomaly detection in large-scale web portals through the CAWAL framework. Knowledge-Based Systems, 306, 112710. DOI: 10.1016/j.knosys.2024.112710