Predicting survival factor following suicide attempt in Iran: an ensemble machine learning technique

Suicide represents a significant challenge to public health that calls for a suitable intervention from the healthcare sector. Despite the typically low suicide rate among most Muslim nations, research indicates that there is an increase in suicide in Iran. Despite increasing suicide rates in Iran, existing predictive models rely on traditional statistical methods, which may be insufficient for individualized risk assessment. This study applies ensemble ML techniques to improve survival prediction accuracy.

Utilizing an extensive dataset collected from a longitudinal study (2017-2024) that includes demographic, psychological, economic, and social factors, we applied several ensemble ML techniques, including AdaBoostM1, J48 pruned tree, Bagging, LogitBoost, MultiBoostAB, J48, SVM, LibLINEAR, and Multilayer Perceptron, to determine critical survival factors after a suicide attempt.

The results reveal that the LogitBoost ensemble models outperformed other algorithms, obtaining an accuracy of 94.3%, with the J48 algorithm following closely at 93.6% accuracy. It is also noteworthy to highlight that the timing of admission is the most influential factor, followed meticulously by the identification of the types of drugs utilized during the suicide attempt.

This study offers important insights into the factors influencing survival after suicide attempts in Iran. It underscores the promising role of ML technique in mental health research. The findings also may guide personalized mediation to enhance support for at-risk populations.