Improving Early Autism Detection with Chi-Square Feature Selection, Machine Learning, and Explainable AI

This study presented a framework that utilized Chi-square feature selection and Machine Learning (ML) classifiers to improve the early detection of Autism Spectrum Disorder (ASD) for children 12 to 36 months old. Six classifiers -Light Gradient Boosting Machine (LGBM), Extra Trees (ET), Decision Tree (DT), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP) -were tested. The findings revealed that the integration of Chi-square feature selection with SVM achieved perfect accuracy, precision, recall, and F1-score, while the other models demonstrated notable gains (up to 90%). Additionally, a SHapley Additive exPlanation (SHAP) analysis was conducted to interpret the model predictions and highlight the key behavioral features, while a literature comparison with recent research showed that the current method outperformed the latter. This study demonstrated that integrating robust feature selection with explainable ML models can significantly advance the reliability of early ASD screening tools.