Improving the Detection Accuracy of Neurological Disorder through EEG-Based Analysis Leveraging Random Forest Classifier for Enhanced Accuracy Compared to K-Nearest Neighbours Classifier

The aim of the proposed research is to enhance the accuracy of the detection of neurological disorders through the use of advanced machine learning on the electroencephalogram (EEG) data. The data in this study was taken on the Kaggle EEG repository which included 8,065 samples in form of structured CSV files. Two popular machine learning to use (Random Forest and K-Nearest Neighbors (KNN) classifiers) were applied and compared regarding the accuracy of detecting the presence of neurological disorders. The random forest classifier showed better performance because of its capacity to deal with high-dimensional, complex features and eliminate overfitting and the KNN classifier provided some extra comparative information though it was lower in performance. The G power analysis was done with a power of 80 and a level of significance (0.05) to guarantee the statistical reliability. Moreover, model performance was checked with the help of confusion matrices, ROC-AUC analysis, 10-fold cross validation. The findings prove that the KNN model is still inferior to the Random Forest classifier, which has much higher classification rates compared to the former with regard to EEGbased diagnostics.