A Correlation-Based Feature Selection and Classification Approach for Autism Spectrum Disorder

Abstract

Autism spectrum disorder (ASD) is a medical condition in which an individual has certain behavior abnormalities, language impairment, and communication problems in the social world. It is a kind of a neurological setback that hinders the ability of an individual. In this work, an effort is made to propose an efficient machine learning-based classifier to assess the individuals on the parameters laid down for ASD based upon the traits captured from the ASD-affected individuals. The standard dataset of 1,054 toddlers is taken, which consists of two categories of toddlers, namely affected by ASD and not affected. The dataset contains 17 features, amongst which 12 features have been selected using correlation-based feature selection, and the random tree classifier gave the best overall performance with an accuracy of 98.9% with 17 features and 99.7% with the selected feature set. The results thus obtained have been compared with other state-of-the-art methods, and the proposed approach outperforms most of them.