Predictive Analysis of Autism Spectrum Disorder using Machine Learning Techniques*

This thesis examines autism spectrum disorder (ASD) forecast with advanced machine learning methods, specifically gradient enhancement models, neural networks, and a hybrid model combining them. The overall goal is to increase prediction accuracy in Identifying ASD, to facilitate early intervention and support. Our experimental findings reveal that by exploiting the strengths of gradient enhancement and neural network, the combined model achieved the best overall accuracy of 86.875% notably that is, both individual models showed adequate cross-validation accuracy, with gradient enhancement 84. 125 % and neural network 85. 125 %. The findings demonstrate how well machine learning techniques work to comprehend the intricacy of 19th-century ASD research data. Additionally, demonstrating the synergistic benefits of model fusion, the combined model performs better than the gradient boosting model alone. Our study confirms that the hybrid approach preserves the robustness of the individual models and improves accuracy is increased. As this model continues to be refined, it is poised to increase its application in ASD research, providing healthcare professionals and caregivers with a powerful tool to identify and support individuals across the autism spectrum. The consistent performance of the combined and individual models further highlights the consistency of the predictive power. This study shows the promise of machine learning integrated approaches in healthcare and offers a strong platform for future research in ASD prediction.