Exploring Machine Learning Models for Predicting Suicide Rates

Psychosocial factors influence suicide and are a major preventable cause of premature death. Mental health disorders are the major cause of self-harm and suicides worldwide. Machine learning techniques could improve methods for assessing suicide or self-attack risk. Information and communication technologies (ICTs) are used considerably in delivering psychological services, and it is an effective way of increasing individual access to mental health diagnosis and treatment. ICT has the potential to access mental health resources for people with disability and consultation with a psychologist. This paper presents contributions focusing on ICT acceptability and psychological services for mental health and exploring machine learning models for predicting suicide rates. The research investigates employed supervised machine learning techniques to predict the suicide rate, namely K-Nearest Neighbours (KNNs), Linear Regression, Decision Trees: Regression, Random Forest: Ensemble of Decision Trees, and XGBoost Regression, with performance evaluation criteria including accuracy and RMSE. It was found that XGBoost was a more effective prediction model. The Chi-square test is performed to see the dependency between the age group and suicide rate. This Chi-square statistic value is greater than the critical value, and the p-value is less than the alpha (0.00003, p ≤ 0.5). Therefore, there is a dependency between the age group and the suicide rate. Irrespective of age group, the male population are more prone to commit suicide than females.