Machine Learning Methodologies for Predicting Autism Spectrum Disorder across Generations

Autism spectrum disorder (ASD) is a neural and developingcircumstance that disturbs children’s social and cognitive skills. It can lead to issues with communication, restricted interests, repetitive behaviors, and trouble in interacting with others. The effect of ASD can be overcome with an initial diagnosis. One of the newest methods for accurately diagnosing ASD in its timely stages or preventing its lasting effects is Federated Learning (FL). This research presents a novel application of the FL approach for autism detection, involving the local training of two distinct machine learning classifiers, namely Self Organizing Map(SOM) and Support Vector Machine (SVM), for the purpose of classifying ASD variables and detecting ASD in both children and adults. Four distinct ASD datasets, consisting of both adults and children, were acquired from various repositories in order to extract features. In children, the suggested typeexpected ASD with $99 \%$ accuracy, and in adults, $83 \%$ accuracy.