Hybrid Ensemble Learning for Autism Spectrum Disorder Detection Using Autoencoders and Classifiers

An early and precise diagnosis is necessary for an effective intervention for autism spectrum disorder (ASD), a neurodevelopmental disorder that affects social communication and behaviour. Convolutional Neural Networks (CNNs) have been investigated for the detection of ASD; however, they frequently have trouble processing structured and unbalanced data. This research suggests a Hybrid Ensemble Learning method that combines: MultiLayer Perceptron for identifying intricate patterns, Autoencoders for unsupervised feature extraction, XGBoost for robust classification, and One-Class SVM for anomaly detection. The framework outperforms CNN-based methods by utilizing the advantages of these models to increase accuracy, sensitivity, and specificity.