AUTISM SPECTRUM DETECTION

Abstract: The increasing prevalence of Autism Spectrum Disorder (ASD) underscores the need for accurate early detection methods to facilitate timely intervention. This study investigates the efficacy of computational models in ASD detection by leveraging both numerical data and image datasets. Employing Support Vector Machine, Logistic Regression, Random Forest, and Neural Network algorithms for numerical data analysis, and utilizing an EfficientNet model for image data analysis, a comprehensive approach is adopted. The numerical dataset, consisting of 2000 samples, yields an accuracy rate of up to 98% with grid search cross-validation using a Decision Tree classifier. Meanwhile, the image dataset, comprising 2500 images, achieves a 94% accuracy rate with the EfficientNet model. By integrating findings from both numerical and image analyses, this study provides a comprehensive report comparing the results and demonstrating the potential of combined approaches in enhancing ASD detection accuracy.

Keywords: CNN (Convolution Neural Networks), Deep learning, SVM (Support Vector Machine), Pre-processing, Feature Extraction, Segmentation.