A Machine Learning Approach for Early-Stage Autism Spectrum Disorders Detection

Abstract

Abstract: The undertaking proposes a Machine Learning (ML) system for early ID of Autism Spectrum illness (ASD), recognizing the hardships of taking out the sickness yet endeavoring to diminish its seriousness through early therapies. The proposed system tests four Feature Scaling (FS) methods (Quantile Transformer, Power Transformer, Normalizer, Max Abs Scaler) on four normal ASD datasets from babies to grown-ups. ML calculations (e.g., Ada Boost, Random Forest, Decision Tree, K-Nearest Neighbors, Gaussian Naïve Bayes, Logistic Regression, SVM, LDA) are utilized on included scaled datasets. The best classifiers and FS strategies for each age bunch are recognized utilizing factual estimations. The voting classifier predicts ASD with the greatest accuracy for Babies, Kids, Young people, and Grown-ups. The task incorporates a definite element significance examination utilizing four Component Determination Strategies to underline the significance of calibrating ML techniques in foreseeing ASD across age gatherings and to assist medical services specialists with pursuing ASD screening choices. In contrast with current early ASD discovery strategies, the proposed structure performs well. To further develop ASD identification strength and precision, a group procedure using a Voting Classifier with Random Forest (RF) and AdaBoost achieved 100% accuracy.