Performance Analysis of Machine Learning Classifiers for ASD Screening

Several machine learning classifiers have been used for Autism Spectrum Disorder screening, however, literature in finding the best classifier for this application domain is inadequate. Hence, this paper presents a comparison of five (5) supervised machine learning algorithms: Decision Tree, Naïve Bayes, k-nn, Random Tree, and Deep Learning using small datasets (n=1100) on child, adolescent and adult ASD screening in finding the most appropriate classifier. These algorithms, which are evaluated using a broad set of prediction performance metrics including accuracy, precision/recall measures, and Receiver Operating Characteristics, are compared against each other. The experiment result suggests that the Deep Learning classifier gives the best performance (with more than 96%) in almost all metrics while the Random Tree classifier came out as the least performing classifier in all the performance metrics.