Identification of ADHD biomarkers by combining ensemble learning with feature weight analysis

This study enhances the classification performance of ADHD (Attention Deficit Hyperactivity Disorder) identification through an ensemble voting method and identifies potential biomarkers using feature weights. Addressing the high dimensionality and complexity of resting-state functional magnetic resonance imaging (fMRI) data in ADHD, we propose an ensemble learning-based framework that combines multiple base classifiers (such as SVM, KNN, and Decision Tree) for voting classification. By optimizing the weights of each classifier, we not only compare the classification accuracy of ADHD under different proportions of the training set but also identify the key features contributing most to the classification results through feature weight analysis. Experimental results demonstrate that this method exhibits excellent classification performance on the ADHD-200 dataset and successfully filters out potential biomarkers associated with ADHD.