An Efficient Predictive Model for Suicide Attempts in Bangladesh Using Machine Learning Algorithms

Nowadays, suicide risk assessment for predicting suicide attempts is time-consuming, has limited predictive value, and lacks dependability. It is a difficult and complex process to predict suicide attempts since it takes many different aspects and things into account. As in other nations, suicide is a major public health concern in Bangladesh. While there isn’t much evidence available on models designed particularly for Bangladesh that predict suicide attempts, there are a number of risk variables that are frequently linked to suicidal behavior. It is important to note that because of the complexity of human behavior and the particular circumstances that surround each person’s life, it is extremely challenging to anticipate individual suicide attempts with high accuracy. Researchers and mental health practitioners are always attempting to create and enhance risk assessment instruments and predictive models that can assist identify those at higher risk. In this study, we attempted to test and train five machine learning algorithms using a dataset of 170 patients diagnosed with various illnesses in Bangladesh. Decision Tree classifier, support vector classifier (SVM), Nave Bayes, Ensemble and CNN are the names of machine learning algorithms. Here the accuracy of the train result of the Decision Tree classifier is much, which is 0.966, but the accuracy of the test result of CNN, Ensemble and SVM is much, which is 0.922 by comparing them. Our discoveries demonstrate that set of experiences of suicide endeavor, Gender, Age, Religion, Occupation, Education, Employment status, Marital status, social class, Depression, Anxiety disorder, Personality disorder, Academic failure, Sexual abuse, Medical problem, Previous Alcohol abuse, suicide ideation and seriousness of clinical discouragement are valuable variables for the expectation of suicide attempts.