Facial Image Detection for Severity Level Prediction of Autism Spectrum Disorder Using Machine Learning Algorithm

Autism spectrum disorder (ASD) is a neurodevelopmental condition impacting social interaction, communication, and behavior. Treatment for ASD is individualized based on three severity levels. Machine Learning streamlines clinical assessments by utilizing facial expression recognition to gauge autism severity, employing distance features processed through Euclidean and Geodesic methods. The proposed machine learning approach for assessing autism severity, incorporating landmark detection via gaze probabilities, has been validated by experts and therapists, achieving an 83% accuracy rate. Based on the experimental results, the Decision Tree exhibits superior accuracy at 83.42%, outperforming KNN (82.58%), Random Forest (82.17%), Gradi- ent Boosting (81.44%), XGBoost (79.04%), Adaboost (78.42%), and Logistic Regression (74.77%) in terms of classification accuracy when predicting facial images.