A Hybrid Approach to Attention Deficit Hyperactivity Disorder Detection Leveraging Transformer and XGBoost Models Using XSparseFormerNet

Early detection of neurodevelopmental disorders such as attention-deficit/hyperactivity disorder (ADHD) is crucial for improved outcomes and prompt intervention. However, traditional detection methods often suffer from challenges due to subjectivity and misinterpretation, lack of resources, and diagnostic biases that can lead to underdiagnosis or overdiagnosis. Early detection of these neurodevelopmental disorders not only helps individuals obtain proper ministrations but can also improve their social, cognitive, and mental development. This study proposes an ensemble model-XSparseFormerNet-that leverages, a custom encoder-decoder Transformer enhanced with various attention mechanisms alongside an XGBoost model to enhance diagnostic accuracy. The proposed model aims to increase the accuracy and efficiency of diagnosis via a custom transformer architecture with the gradient boosting algorithm XGBoost. By using a preprocessed EEG dataset and a customized ensemble model, XSparseFormerNet achieves 85% accuracy, outperforming traditional methods across evaluation metrics. Moreover, this research contributes to future development in the field by offering methodologies that can be useful for studies on disorder detection.