Predictions of Academic Performance of Children and Adolescents with ADHD Using the SHAP Approach.

Abstract

Attention-Deficit/Hyperactivity Disorder (ADHD) is a neuro-developmental disorder characterized by inattention and/or impulsivity-hyperactivity symptoms. Through Machine Learning methods and the SHAP approach, this work aims to discover which features have the most significant impact on the students' performance with ADHD in arithmetic, writing and reading. The SHAP allowed us to deepen the model's understanding and identify the most relevant features for academic performance. The experiments indicated that the Raven_Z IQ test score is the factor with the most significant impact on academic performance in all disciplines. Then, the mother's schooling, being from a private school, and the student's social class were the most frequently highlighted features. In all disciplines, the student having ADHD emerged as an important feature with a negative impact but less relevance than the previous features.