UPSI Digital Repository (UDRep)
|
UPSI Digital Repository (UDRep)
|
|
|
|
||||||||||||||||||||||||
| Abstract : Perpustakaan Tuanku Bainun |
| Violence against women (VAW) is an urgent problem that not only has severe physical impacts but also psychological impacts. Predicting violence based on its type increases efficiency, accuracy, and early detection. This ensures that prevention, intervention, and policies can be more targeted and focused, as interventions for each type of violence vary and require a tailored approach. This study aims to develop the latest prediction model to predict VAW in each type which were physical, psychological, economic, and sexual violence. The methodology used is machine learning with the naive bayes, random forest, and logistic regression algorithms because it provides a balance of efficiency, ease of interpretation, and the ability to handle complex data often encountered in the context of predicting VAW. A representative research sample of 600 married women, data collection through questionnaires designed to capture variables and characteristics associated with various types of VAW. The design of the prediction model in this study is very important for each type of violence because interventions for victims vary in their needs and challenges. Therefore, these findings have significant value in making precise and accurate predictions. Random forest had the best model with 98.75% accuracy, 99.4% AUC, 97.88% precision, 97.25% sensitivity, and 96.99% specificity. Furthermore, when considering VAW by type_physical violence, sexual violence, economic violence_the random forest algorithm performed best with accuracies of 93.33%, 90.00%, 85.83%, and the prediction of psychological violence, the logistic regression algorithm achieved the highest accuracy of 81.67%. This research contributes to improving predictive models for identifying potential risks of violence, developing datasets, and supporting policymakers with data-driven insights. These insights inform the design of a new model optimized to predict different types of VAW more accurately. Ultimately, this contributes to the development of more effective prevention and intervention strategies according to the type of violence. |
| This material may be protected under Copyright Act which governs the making of photocopies or reproductions of copyrighted materials. You may use the digitized material for private study, scholarship, or research. |