AI-Driven Autism Spectrum Disorder Screening Using LightGBM and Behavioral Analysis
Abstract
Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by problems with communication, social interaction, and behavior patterns. Early detection of autism spectrum disorder is crucial for appropriate treatment and positive development. The existing procedures are lengthy and rely on manual assessment. This research work proposes a machine learning solution for the early detection of autism spectrum disorder based on the behavioral data obtained from a questionnaire, along with the demographic features. Different approaches to classification, including Decision Tree, K-Nearest Neighbors (KNN), and Light Gradient Boosting Machine (LightGBM) are tested to select the most efficient method for prediction. LightGBM proves to be the best approach because of its efficiency with structured data. In order to improve model transparency and facilitate decision-making in clinical practice, SHapley Additive exPlanations (SHAP) are incorporated to assess and measure the impact of specific features. The developed model ensures 97% accuracy, 97% precision, 92% recall, and 94% F1-score on the test dataset. The feature importance analysis reveals that the behavioral factors such as communication issues, problems in making eye contact, interaction problems, and repetition are among the important features contributing to the prediction of ASD. By combining high predictive accuracy with transparency, the proposed solution presents a reliable, efficient, and easy-to-use approach for early ASD screening.