SHAP-Driven Interpretability of Autism Risk in Pregnancy Using Explainable AI

Explainable artificial intelligence (XAI) has gained growing popularity for its ability to explain how deep learning and machine learning models make decisions. The frameworks for SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) have become interpretive tools for ML models. This research closely examines the application of LIME and SHAP in the interpretation of autism spectrum disorder (ASD) detection. It stresses XAI's important role in making AI-based ASD predictions more accurate. Researchers also identified risk factors related to ASD by analyzing the impact of features on it. SHAP shows the TG, AGE, and LDL emerge as the primary features contributing to the prediction of ASD. LIME predicts each patient with 55% confidence in having ASD. The study's findings suggest that machine learning approaches can offer accurate forecasts of ASD status, with the suggested model capable of diagnosing it in its early phases.