Deep Dive into Predictive Modeling for Autism Spectrum Disorder using Facial Features and Machine Learning

Autism Spectrum Disorder is a neuro-development behavioural disorder that is classified as a chronic disability leaving individuals unable to accurately grasp their environment and can lead to limited development of communication skills, thinking and other social skills. Prediction of ASD using video recordings is an open-ended critical problem in machine learning to facilitate the process of early intervention which requires high computational resources. This research aims to predict ASD in children using videos by extracting features and converting them into a binary classification problem thereby eliminating the need for high computational resources. This research postulates a study of machine learning techniques like Logistic Regression, Decision Trees and Neural Networks in comparison for the prediction of ASD as well as a novel approach to reduce the computational requirements for the analysis of videos. This research also addresses the lack of publically available video datasets for the analysis of Autism. The highest accuracy metric for the models was using the Neural Network architecture closely followed by the Decision Tree Classifier which yielded accuracies of 99.8% and 99.65% respectively. In this article, we offer an overview of predicting Autism Spectrum Disorder whilst focusing on the importance of video datasets.