Optimized Machine Learning Techniques for Accurate Autism Spectrum Disorder Diagnosis

Abstract

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by deficits in communication, social interaction, and behavior. Early diagnosis and prompt intervention can help to maximize developmental outcomes. This work proposes a refined predictive ASD model using other advanced techniques of machine learning. We leverage two datasets—“Autism Spectrum Disorder Screening Data for Toddlers in Saudi Arabia” and the “Autism Prediction Dataset”—to develop a robust predictive model. The methodology includes extensive data preprocessing, feature engineering, and algorithm selection to improve classification performance. Various machine learning models, including Random Forest, Support Vector Machine (SVM), Gradient Boosting, and Neural Networks, are employed and evaluated based on accuracy, precision, recall, and F1-score. The study also integrates explainable AI (XAI) techniques to interpret model predictions, enhancing transparency in decision-making. Experimental results demonstrate that the proposed approach significantly improves the predictive accuracy of ASD detection compared to traditional methods. The findings suggest that AI-driven diagnostic tools can assist healthcare professionals in making more informed decisions, ultimately aiding in early intervention strategies. Future work includes expanding the dataset and refining models for real-world applications.