Classification of Attention Deficit Hyperactivity Disorder Using Machine Learning

Abstract

The global burden of mental ailments continues to rise, posing serious health risks as well as huge social, human rights, and economic ramifications in every country. One such disorder is the attention deficit hyperactivity disorder (ADHD) which is prevalent among children and teenagers. There is no single test that can diagnose ADHD. Symptoms must pose problems in at minimum two places (such as school, family, job, or leisure time) for at least six months to be diagnosed. Facing issues in paying attention by youngsters can lead to low academic performance. In addition to this, ADHD is sometimes linked to various mental illnesses and substance abuse issues, which can lead to further harm, especially in the contemporary generation. Unfortunately, ADHD is incurable. But early detection, together with an effective treatment and education plan, can help a child or adult with ADHD manage their symptoms. Therefore, this project attempts to classify ADHD using machine learning (ML) techniques in order to help provide valuable insights on establishing an automated diagnosis system. A comparative analysis of 4-way classification of ADHD using various machine learning algorithms has been done in WEKA toolkit (experimenter) while also experimenting with different subsets of features, including those created by applying genetic algorithm (GA), from the phenotypic characteristics of the ADHD-200 data set. The ML classifiers that have been used are Logistic, Support Vector Machine (SVM), Decision Tree (DT); implemented through J48 algorithm, Random Forest (RF), K-nearest neighbour (KNN); implemented through the instance-based learner (IBk) algorithm, and multi-layer perceptron (MLP). A total of 8 performance parameters were used for the evaluation of these classifiers: accuracy, precision, recall, F-measure, Kappa-statistic, root mean squared error (RMSE), Mathew's correlation coefficient (MCC), and area under the receiver operating characteristics (AUROC) curve.