Victim Decision-Making Models: Rationales for Justifying Victims’ Non-Reporting of Stalking – A Machine Learning Approach

• In Event: Computational Methods in Criminal Justice Settings

Abstract

This study examines victim decision-making models to understand the rationales behind stalking victims’ non-reporting behaviors using a machine learning approach. Utilizing data from the 2019 Supplemental Victimization Survey (SVS), which includes 1,406 stalking victims, we analyze key factors—such as victim-offender relationships, financial loss, physical threats, and demographic characteristics—that influence reporting decisions. Machine learning techniques, including Logistic Regression, Decision Trees, Naïve Bayes, Artificial Neural Networks (ANN), and K-Nearest Neighbors (KNN), are applied to identify distinct subgroups of non-reporting victims. Dimensionality reduction (PCA), feature selection, and normalization enhance model performance, while Ridge and Lasso regression with k-fold cross-validation ensures generalizability. By uncovering patterns in victim decision-making, this study provides empirical insights into the barriers to reporting stalking incidents and supports the development of targeted interventions to improve criminal justice responses and victim support services.

Authors

• Sujung Cho, Southern Illinois University Carbondale

• Rasheed B. Ibrahim, Southern Illinois University Carbondale