Statistical Comparison of Some Machine Learning Techniques: A Case Study for Classifying Domestic Violence Crimes in Iraq

Abstract

Crime classification and prediction represents a contemporary societal trend aimed at reducing or preventing criminal activities. One such crime is domestic violence, a global phenomenon known as the shadow pandemic or the hidden pandemic. This study aims to analyse data related to domestic violence crimes in Iraq and compare the effectiveness of the Random Forest (RF), Decision Tree (DT) and Naïve Bayes (NB) models using different metrics to develop an accurate model for classifying or predicting the type of domestic violence (physical or non-physical). The research employs a methodological framework that includes several stages. Subsequently, the three classifiers RF, DT, and NB are applied to the dataset to facilitate classification and prediction. The experiment results showed the superior performance of the RF classifier, achieving an accuracy score of (99.77%), compared to the DT classifier (99.07%) and NB (97.69%). For validation, different classification metrics were used. RF exhibited superior performance in all metrics compared to DT and NB algorithms, whose performance capabilities varied.