Help-seeking behaviors of stalking victims: Integrating machine learning and regression approaches to examine how victimization consequences shape victims' decisions

We examined how the consequences of stalking victimization shape victims' help-seeking behaviors, using the 2019 National Crime Victimization Survey, Supplemental Victimization data. We analyzed three distinct help-seeking outcomes: (a) reporting to police, (b) help-seeking from victim-serving agencies, and (c) help-seeking from personal networks. Logistic regression models assessed the statistical significance of individual predictors, while our non-parametric Machine Learning approach evaluated their predictive power and captured non-linear patterns. Regression results revealed that substantial emotional distress significantly increased the likelihood of all three help-seeking behaviors. Health, social, and financial problems increased the likelihood of network help-seeking, while social problems were associated with lower odds of police reporting. Machine learning identified financial problems, emotional distress, and health problems as the most predictive features for police reporting, agency help-seeking, and network help-seeking, respectively. These findings underscore the multidimensional nature of victimization consequences and the value of combining traditional statistical inference with machine learning to better understand victim decision-making.