An Experiment on Logistic Regression Analysis to Detect Autism Spectrum Disorder

Abstract

Autism is a behavioural disorder that commonly affects the communication and interaction of an individual. Early detection may deduce symptoms and support the daily living of a person with the assistance of therapy. Many researchers investigate the factors associated with autistic traits that pro-duce meaningful results for further analysis. In this paper, an experiment on one such analysis is focussed on a toddler, child, adolescent and adult autism dataset using a machine learning technique, specifically logistic regression. In the analysis, feature selection techniques applied are chi-square and information gain to reduce the dimensionality of the dataset. The experimental analysis results with a mean accuracy of 90 %. In addition, a few hypotheses are proposed with the evidence obtained from the dataset as an initial step of the research process.