A Proposal of Data Mining Model for the Classification of an Act of Violence as a Case of Attempted Femicide in the Peruvian Scope

Abstract

Nowadays, femicide is one of the biggest problems worldwide in which the human rights of the victims are violated. In addition, it also constitutes a public health concern, with serious physical and psychological consequences. The objective of this research is to implement a data mining model to classify an act of violence as a case of attempted femicide in Peru. This study used public data of 2021 of the statistics portal National Aurora Program of the Ministry of Women and Vulnerable Populations (MIMP). The applied methodology was based on 5 phases: Data collection, data understanding, data preprocessing, data mining and model evaluation. Results obtained with Balanced Random Forest and Logistic Regression models demonstrated the best performances with a Recall of 0.88 and 0.86, respectively. Furthermore, the application of SMOTE improved the performance of both models. This investigation will contribute to find patterns related to the characteristics of aggressors and victims, that can help to put into action new instruments based on Data Mining to prevent more murders of women.KeywordsAttempted FemicideBalanced Random ForestClassification ModelData MiningFeature selectionLight GBMLogistic regressionSMOTE