An Interpretable Classification Model for Identifying Individuals with Attention Defict Hyperactivity Disorder

Abstract

Attention Deficit Hyperactivity Disorder (ADHD) is a psychiatric condition that affects around 5% of children around the world. The primary attention procedure is traditionally based on analysis of ratings collected in questionnaires called psychometrics. This work aims to investigate interpretable classification models capable of not only accurately identifying individuals with ADHD, but also explain it, by providing the evidences that lead to the outcome. We compare the performance of Explainable Boosting Machine (EBM) with 3 other classical decision tree-based models and observed similar results, with the distinction of EBM being a more interpretable model. We also assess explanations quantitative and qualitatively, demonstrating how they may actually help psychiatrists in their practice.