Exploration of sexual violence in the city of Bogota: application of a data mining technique

Abstract

This study offers an approach to sexual violence by using secondary source data and applying some data mining techniques. The data source used is the 'Instituto Nacional de Medicina Legal y Ciencias Forenses' (National Institute of Legal Medicine and Forensic Sciences), and the algorithms applied are Selection by Characteristics, C5.0, and K-Means.

Prior to applying these techniques, a theoretical approximation to sexual violence is made in order to appreciate how this kind of violence has been approached and analyzed. Subsequently, data quality is assessed and some improvement treatments are applied. Upon having reached a proper dataset for processing and analysis, data mining techniques are applied, and the relationship of the alleged aggressor to the victim is established as a variable objective or answer.

The issues or solutions offered by the above data processing lead to an analysis which establishes as a core the levels of proximity with the victim, and questions those studies based on the traditional kinship structure, while it simultaneously validates the distinction that establishes a sexual violence rating between abuse and assault. Analyses of the data mining exercise facilitate a clear statement of the configuration of two clusters that can be pointed at with said classification. They are accompanied by a third one that, although not well defined yet, begins to appear. The three clusters have been designated as sexual violence in an incest situation, sexual violence in an anonymity situation, and sexual violence in a family structure situation. Finally, some suggestions are given in seeking to improve data quality, while the opportunities this type of analysis opens at attempting to give an answer to conflictivity, violence and crime are outlined.