Application of machine learning in predicting Attention Deficit Hyperactive Disorder (ADHD) in school going children of Pakistan

Aims: The purpose of this study was to use machine learning algorithms to predict the probability of a child to have a certain attention deficit hyperactive disorder (ADHD) score under a given set of conditions. Methods: This was a cross-sectional survey which employed non-probability convenient sampling technique conducted at two schools in Islamabad, Pakistan. Using the latest version of Konstanz Information Miner (KNIME) Analytics, several machine learning algorithms were tested. Results: The area under the curve (AUC) for classification tree was 60.8% with a precision of 75.6% for the prediction of an ADHD score of 20 or more and the probability of 21.3% for a child to have an ADHD score of 20 or more. Important variables associated with a higher ADHD score included father’s profession, school of the child, and the class of child. Conclusion: This study shows that machine learning approach may be useful in developing a robust predictive model. Use of predictive model may allow use of limited resources towards assessment of children with higher probability of ADHD.