Accurate Classification of Autism Spectrum Disorder Using Stacked Ensemble Learning Techniques

Autism Spectrum Disorder (ASD) is a developmental disorder marked by challenges in social interaction, communication difficulties, and repetitive behaviours. Early and precise identification of ASD is crucial for implementing timely interventions and providing appropriate support to individuals. However, conventional diagnostic approaches are often time-intensive and reliant on subjective clinical evaluations. In recent years, artificial intelligence has emerged as a prevailing tool for improving diagnostic precision by identifying patterns in behavioural, clinical, and demographic data. In this research, a data-driven methodology to classify individuals with ASD using resting-state fMRI data from the ABIDE dataset. To perform the classification, Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and Neural Networks techniques have been applied. Furthermore, a stacked model that leverages fine-tuned classifiers as base models. However, to further enhance both the performance and interpretability of the classification process, future research should explore alternative functional connectivity extraction methods, moving beyond traditional correlation-based approaches. Such innovations could lead to better model accuracy and contribute to a deeper understanding of ASD, which may aid in earlier diagnosis and more individualized treatment approaches. Experimental results indicate that the proposed model surpasses individual algorithms, highlighting its potential as an efficient and reliable tool for ASD detection.