A Machine Learning Approach for Autism Spectrum Disorder Detection Using Behavioural Data and Facial Features.

Autism Spectrum Disorder (ASD) is a developmental condition that influences how a person communicates, behaves, and relates to others. Detecting ASD as early as possible is very important, since early therapy and support can improve long-term outcomes. In this study, we propose a dual approach that combines behavioral data with facial image analysis to improve detection accuracy. For behavioral data, we used the Q-CHAT-10 questionnaire and applied the Random Forest Classifier (RFC), which is known for handling complex patterns effectively. For image data, we applied the YOLOv5 model to detect faces and then classified the extracted features using a Support Vector Machine (SVM). Both datasets used in this work were publicly available and carefully preprocessed to ensure data quality. The behavioral model achieved 98% accuracy, while the image-based YOLOv5+SVM approach reached a mean average precision (mAP) of 0.91. When results from both models were combined, the system achieved an overall accuracy above 97%. These results suggest that integrating behavioral screening with facial analysis can provide a faster, more reliable tool for early ASD detection, supporting timely intervention and personalized care.