Swarm Intelligence and Deep Learning based Approach for the Early Prediction of Autism Spectrum Disorder among Adults

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that affects how people communicate and shows itself through behavioral expressions and social interaction deficits. Successful interventions together with proper support require an early and precise diagnosis of ASD in adult individuals. The proposed research presents a deep learning based prediction system for ASD which combines Bird Swarm Optimization (BSO) framework for both selecting features and optimizing model parameters. A publicly accessible ASD dataset obtained from Kaggle repository serves as the training base for the model where behavioral and psychological and demographic information was obtained from standardized autism diagnostic data. The predictive model benefits from BSO because it chooses the significant attributes while optimizing deep neural network (DNN) hyperparameters. Results from experiments proved that the optimized model exhibits better accuracy in addition to precision and recall compared to conventional machine learning and deep learning approaches. Results of research further indicated that the BSO-enhanced deep learning model functions successfully as an ASD screening method for adults which helps practitioners make early diagnoses followed by customized treatment strategies.