Performance Analysis of Machine Learning Techniques to Predict Mental Health Disorders in Children

Mental disorders are quite common in children. The commonly found childhood mental disorders are anxiety disorders; depression and attention deficit disorder. Diagnosis of these problems at early stage helps the professionals in treating it at beginning stage and to improve the patient’s health. Therefore the need to treat common mental health disorders that are found in children which lead to complicated problems, if ignored at early stage. Machine learning Techniques can be applied for analyzing patient’s history to diagnose the problem. In this research three machine learning techniques have been identified and compared based on their performances on several scales of accuracy on selected attributes to diagnose five basic mental health disorders. The basic aim is to find the technique which is most accurate. The data set is containing sixty attributes for analyzing and measuring the performance of techniques. Ignoring the irrelevant attributes that do not have much effect, twenty-five attributes were found as important to diagnose the disorder. Applying Feature Selection algorithms on the attribute set, thirteen attributes were found. Accuracy of the selected attribute set on three machine learning techniques were compared viz., Multilayer Perception, LAD Tree and Multiclass Classifier. It is clear by the results that the Multiclass classifier produces much accurate results on set of selected attributes.