A Machine Learning Approach to Predict Poor Mental Health of Intimate Partner Violence Survivors

Abstract

Intimate Partner Violence (IPV) is a wide social problem in Canada and abroad. Survivors of IPV are likely to experience mental health challenges. Detecting the experience of mental health challenges is paramount to address them as early as possible. Using a Statistic Canada survey (General Health survey, 2014), we have built a machine learning approach to predict the experience of poor mental health among IPV survivors. Multi-Layer Perceptron (MLP) provide the best accuracy score of 94.88 for a 14-feature model, and 94.21 % for a 24-feature model. The use of a more detailed dataset from Statistics Canada is recommended. Multidisciplinary research has a great potential in this emerging field.