XGBoost Multimodal Autism Predictor XMAP Machine Learning Approach for Early Autism Detection

This study introduces the XGBoost Multimodal Autism Predictor (XMAP), an interpretable machine learning framework designed to improve ASD classification by integrating publicly available behavioral datasets with synthetically generated features. Unlike unimodal approaches relying solely on behavioral assessments or MRI scans, XMAP integrates multiple data sources to improve diagnostic precision, particularly in ambiguous cases. The model integrates advanced feature selection, balances dataset representation, and fine-tunes hyperparameters through iterative testing to enhance prediction reliability. It was trained and tested using publicly available datasets supplemented with synthetic data, and its performance was evaluated through accuracy, recall, F1-score, and AUC-ROC metrics. Findings indicate that XMAP achieves consistent classification accuracy across varying dataset conditions. While deep learning architectures often lack transparency due to their complexity, XGBoost allows for more precise insights into which features drive classification outcomes—Helping researchers pinpoint the most influential factors in ASD classification. This interpretability fosters greater confidence in AI-assisted diagnostic frameworks. Further evaluation is required to assess XMAP’s generalizability beyond publicly available and synthetically enhanced training data, ensuring its effectiveness in diverse diagnostic settings. Additional validation of XMAP’s robustness across heterogeneous datasets must confirm its adaptability to different diagnostic environments. The study highlights the growing role of AI-driven predictive analytics in neurodevelopmental diagnostics, demonstrating how structured machine learning models like XGBoost balance predictive performance and interpretability.