BIOMARKER-DRIVEN PREDICTION OF AUTISM SPECTRUM DISORDER USING ENSEMBLE LEARNING APPROACHES

Abstract

Background: Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition characterized by impairments in social communication, repetitive behaviors, and restricted interests. Early and accurate diagnosis remains challenging due to the heterogeneity of symptoms and the absence of a single definitive diagnostic test. Recent advances in biomarker research and artificial intelligence have opened new avenues for improving ASD prediction and diagnosis.

Objective: This study aims to develop a biomarker-driven predictive framework for Autism Spectrum Disorder using ensemble learning techniques and to evaluate the effectiveness of combining multiple machine learning models for enhanced diagnostic accuracy.

Methods: A dataset comprising biological, genetic, neurophysiological, and clinical biomarkers associated with ASD was analyzed. Data preprocessing, feature selection, and normalization techniques were applied to improve model performance. Multiple ensemble learning algorithms, including Random Forest, Gradient Boosting, AdaBoost, and Extreme Gradient Boosting (XGBoost), were implemented and compared. Model performance was assessed using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC-ROC).

Results: The ensemble learning models demonstrated superior predictive performance compared to individual machine learning classifiers. Among the evaluated approaches, XGBoost achieved the highest classification accuracy and robustness in identifying ASD-related patterns from multidimensional biomarker data. Feature importance analysis revealed that specific biological and neurodevelopmental markers contributed significantly to prediction outcomes.

Conclusion: Biomarker-driven ensemble learning approaches offer a promising strategy for the early prediction and diagnosis of Autism Spectrum Disorder. The integration of advanced machine learning techniques with biomarker analysis can enhance clinical decision-making, facilitate early intervention, and contribute to the development of personalized treatment strategies for individuals with ASD.