Crime Behavior Prediction Using Process Mining and Deep Learning: Toward a Scalable Script Analysis Framework

• Myeonggi Hong,
• Eunbi Cho,
• JeongHyeon Chang,

Vol. 19, No. 9, September 30, 2025
10.3837/tiis.2025.09.018, Download Paper (Free):

• Crime script
• Environmental criminology
• Image Analysis
• Process Mining
• Convergence

Abstract

This study presents an integrated approach to crime script analysis by combining criminological theory with advanced computer science technologies. Crime scripts conceptualize criminal behavior as a structured sequence of actions, serving as a foundation for analyzing recurring crime patterns and supporting proactive crime prevention strategies. Despite their theoretical utility, traditional crime analysis methods often suffer from limitations in data scalability, dynamic model updating, and the integration of heterogeneous data sources. To overcome these challenges, this paper critically reviews existing crime script methodologies and proposes a novel analytical framework that leverages deep learning techniques. The proposed method enables automated extraction, classification, and prediction of crime sequences from large-scale behavioral datasets. Methodological case studies are included to demonstrate the operational feasibility and logical coherence of the framework across various crime scenarios. The proposed framework processes video data nearly twenty times faster than traditional methods. It reduces researcher bias through automated coding and enhances reproducibility with quantitative process mining. This study establishes a scalable, objective methodology that bridges criminological theory and artificial intelligence. The goal of this study is to establish a flexible and scalable crime analysis architecture that enhances the effectiveness of crime prevention through data-driven, adaptive modeling. In doing so, this research contributes to the advancement of computational criminology and offers practical implications for the development of intelligent law enforcement support systems.

Statistics

Cite this article