Image Classification of Autism Spectrum Disorder Children Using Naïve Bayes Method With Hog Feature Extraction

Abstract

Autism Spectrum Disorder (ASD) is a developmental disorder that affects a person's ability to communicate and interact socially. Every year, the number of people diagnosed with Autism Spectrum Disorder rises, necessitating early detection in order to limit the number of people affected and provide proper treatment. As a result, a system was developed in this study to detect Autism Spectrum Disorder in facial photos utilizing versions of the Nave Bayes approach and HoG feature extraction. HoG feature extraction is a local intensity gradient distribution or edge direction perpendicular to the gradient direction without influencing the geometric and photometric transformations, and Nave Bayes is a method that classifies images based on probability. The experimental results of three types of naive Bayes, Bernoulli naive Bayes is the most reliable than Multinomial naive Bayes and Gaussian Naive Bayes. Accuracy, Precision, Recall, and the highest F1-Score using this method, with each value of 89.72%; 90.54%; 89.72%; and 89.9%. The next best performing Gaussian Naive Bayes, the most laborious results were obtained using Naive Bayes multinomials, which had Accuracy, Precision, Recall, and F1-Score of 65.91% each; 68.09%; 65.91%, and 64.19%.