Use of Artificial Neural Networks in Determining Domestic Violence Predictors

• Published in: Advances in Swarm Intelligence Vol. 11656; pp. 132 - 141
• Main Authors: Silva, Jesús, Aleman, Erick Guerra, Acuña, Génesis Camargo, Bilbao, Osman Redondo, Hernandez-P, Hugo, Castro, Bella León, Meléndez, Pedro Arrieta, Neira, Dionicio
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2019; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Artificial neural network; Domestic violence; Multilayer perceptron
• ISBN: 3030263533; 9783030263539
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-26354-6_13

This paper aims to determine the predictors of violence against women by their partners, according to the National Survey of Demography and Health - ENDS - Colombia, 2017 using artificial neural networks. The results indicate that the best forecasting model found is the artificial neural network, perceptron model, multilayer associative memory with a hidden layer of 20 units, through functions of sigmoidal activation and sum of square of the error as error function. The ten main explanatory variables are: respect for human rights of the partner, respect for wishes, love expressed by the partner, a history of domestic violence, engaging in joint decision making, decision of contraceptive use, number of connections (partners) of the respondent, decision-making at the financial level, correction of children behavior, and decisions regarding women’s health at home.