A Machine Learning Framework for Identifying Autism Spectrum Disorders in Their Early Stages

Abstract

The project suggests a Machine Learning (ML) approach for early identification of autism spectrum disorder (ASD), acknowledging the difficulties in curing the condition while attempting to lessen its severity with early interventions. Using four typical ASD datasets, ranging in age from infants to adults, the suggested system evaluates four Feature Scaling (FS) techniques: Quantile Transformer, Power Transformer, Normalizer, and Max Abs Scaler. On the scaled datasets that are included, machine learning algorithms (ML) such as Ada Boost, Random Forest, Decision Tree, K-Nearest Neighbors, Gaussian Naïve Bayes, Logistic Regression, SVM, and LDA are used. Based on factual estimations, the optimal classifiers and FS techniques for each age group are identified. The voting classifier is most accurate in predicting ASD in infants, children, adolescents, and adults. Using four Component Determination Strategies, the job includes a definite element significance analysis to highlight the importance of calibrating machine learning algorithms in predicting ASD across age groups and to support medical care professionals in making ASD screening decisions. Compared to existing early ASD finding methodologies, the suggested structure has superior performance. In order to enhance the strength and precision of ASD recognition, a group process that employed a Voting Classifier with Random Forest (RF) and AdaBoost yielded 100% accuracy.