Predicting suicidal behavior: a machine learning model

OBJECTIVE: Timely response and identification of risk factors are critical components of suicide prevention, enabling appropriate intervention. By applying machine learning to clinical data analysis, predictive models can be developed to classify patients into high- and low-risk groups, which can help prevent loss of life. This study aims to develop a machine learning model to categorize patients based on their risk of suicide attempts. MATERIALS AND METHODS: We analyzed data from 301 respondents hospitalized at the Clinic of Psychiatry for either suicidal behavior or non-suicidal reasons. They were divided into two groups according to the presence of suicidal behavior. By using machine learning methods, we analyzed the influence of 18 observed features on the development of suicidal behavior. A predictive model was developed to classify individuals into two risk categories - high and low - for the observed population. The k-Nearest Neighbors (kNN) algorithm was used to train the model. RESULTS: The kNN-based model achieved a classification accuracy of 87%, with sensitivity at 87%, precision at 90% and an F-score of 85% for the tested sample. CONCLUSIONS: Early identification of risk factors for suicidal behavior remains the most effective strategy for prevention. Classification models may serve as valuable clinical tools for assessing suicide risk, potentially contributing to a reduction in suicide rates in Vojvodina. Expanding the database could improve the functions of the obtained classifier and enable its introduction into medical practice.

https://www.europeanreview.org/wp/wp-content/uploads/Graphical-Abstract-27.jpg.