Prediction of Suicidal Risk Using Machine Learning Models

Abstract

According to WHO (World Health Organization), every 40 seconds a person dies of suicide. This amounts to a total of 800,000 people every year falling victim to suicides. Suicide is a global phenomenon: it accounts for 1.4% of all deaths worldwide and costs about $51 billion annually to the healthcare industry. Targeted and timely interventions are critical to helping the patients who are dealing with suicidal symptoms. Data availability is high in the healthcare industry, and this can be used to extract knowledge for better prognosis, diagnosis, treatment, and drug development. In this paper, we have focused on predicting suicidal risk by using various types of machine learning models. The highest accuracy among our predictive machine learning models is 98.8% test accuracy and 96.3% 10-fold cross-validation accuracy using the XGBoost model, which is good compared to existing models present in the literature.

Note:

Funding Information: This research received no external funding.

Conflict of Interests: The authors declare no conflict of interest.

Keywords: Decision Tree, Gradient Boosting, K-Nearest Neighbor, Linear Regression, Machine learning, Multilayer Perceptrons, Random Forest, Suicide, XGBoost