A Hybrid Machine Learning and Regression Approach for Validating a Women’s Safety Crime Index

COMPUTER APPLICATIONS

Violence against women has existed throughout history, taking many forms, from emotional and psychological abuse to physical and sexual harm. This study proposes a crime index that captures factors influencing such violence, organized into four key areas: Health, Socioeconomic Status, Education, and the Judiciary. The index was evaluated for consistency and reliability, and its effectiveness was determined by comparing it with both traditional and combined modeling approaches. An integrated modeling strategy, which used regression techniques alongside Random Forest and Stochastic Gradient Descent, was applied to enhance prediction performance. Model accuracy was evaluated using MAE, RMSE, and MAPE. Results show that the index is highly reliable, with strong correlations. This approach effectively accounts for variation and reduces prediction inaccuracies compared to baseline models, while the homogeneous ensemble achieved the lowest inaccuracies. Women in underdeveloped or geographically challenging areas are more vulnerable to victimization. Overall, the index highlights the important role of social factors in violence against women and can support strategies to reduce such incidents.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.