A Comparative Study of Machine Learning Algorithms in Predicting Mental Disorders

Mental disorders include a range of psychiatric diseases, involving stress, trauma, post-traumatic stress disorder (PTSD), neurocognitive disorders, panic disorder, obsessive compulsive disorder (OCD), autism spectrum disorder (ASD), substance-addictive disorder, bipolar disorder, neurodevelopmental disorder, schizophrenia, and sexual dysfunctions. These are related to unfavorable health behaviors like disturbed sleeping patterns, poor diet, poverty, unemployment, some unusual life events, and higher addiction rates. Of late, depression, anxiety, and bipolar disorders are ruling the world and have become more prominent among the youth belonging to high-income nations. Depression is the most common disease in clinical practice, leading to occupational, social, interpersonal, and financial problems. However, Machine learning (ML) is becoming the main cause of the Big Data insurgence. It has captured applications in diverse domains, for instance, industries, pharmacy, medicinal chemistry, waste treatment plants, social networking, management, robotics, digital advertisements, physiological signal monitoring, electronic health record (EHR), medical & pathological imaging, and smart sensors. These techniques are computer-based equipped on automation, development, adaptation, prediction, and interpretation of the result by the machine. In this context, this review pertains the recent literature information about the real time prediction of various types of mental disorders using machine learning approach. Adaboost predicted depression with $\mathbf{9 2. 5} \%$ accuracy and 93.6 % specificity. Moreover, XGBoost was applied for post stroke depression. Otherwise RNN could be utilized depression from EEG brain wave data. While SVM was explored for androgenic alopecia, However, RF model was evaluated for anxiety.