Predicting Suicide Cases Using Deep Neural Network

Abstract

Suicide is a critical issue in contemporary society, giving rise to significant societal and economic ramifications. To mitigate these adverse effects, it is imperative to implement effective suicide prevention strategies. In this context, deep neural network (DNN) algorithms have gained prominence and are increasingly applied across various healthcare domains. In our research, we examined the efficacy of employing DNNs for predicting suicide attempts. Our study involved a descriptive-analytical, cross-sectional investigation that harnessed a DNN algorithm, specifically a Sequential model comprising four Dense layers, to analyze suicide-related data. We identified several crucial risk factors, such as a history of psychiatric hospitalization, the day of the week, and occupation, for predicting suicide attempts. Leveraging the DNN algorithm on a dataset encompassing 1453 individuals with a suicide history, we framed the problem as a binary classification task with suicide history as the target variable. To enhance model performance, we applied preprocessing techniques, including class balancing, and constructed a DNN model with a Sequential architecture featuring four Dense layers. It is worth emphasizing that the DNN algorithm’s effectiveness hinges on the quality and quantity of available data and the model’s architectural choices. Nevertheless, our study demonstrated satisfactory overall performance of the DNN algorithm. For instance, the model achieved an f1-score of 91%, signifying its high accuracy in predicting positive cases (in this context, suicide incidents) while maintaining a balanced trade-off between precision and recall.