الفهرس | Only 14 pages are availabe for public view |
Abstract A major objective of this thesis is to enhance the classification of Brain Computer Interface (BCI) datasets. Enhanced classification performance is achieved with SBT–RF algorithms using two analyzed stages in this study. First, using terminals, electrodes, and biosensors, Brain Computer Interface (BCI) interacts with medical equipment and computers. Also, human intentions and thoughts are analyzed and recognized using BCI, which is then translated into Electroencephalogram (EEG) signals. Most of the time, researchers use different combinations of classifiers (i.e., bagging and adding kernel) to enhance the accuracy of the classification algorithms. Nevertheless, certain brain signals might contain redundant data, which results in inefficient classification. Thus, feature selection methods have been employed to eliminate redundant data before the classification is performed to decrease computation time. An SBT-RF technique is presented to determine the importance of the features for selecting and classifying the data. SBT-RF is better at improving the mean accuracy of the dataset, it also decreases computation cost, and training time, and increases prediction speed. RF is better at improving the mean accuracy of the dataset, it also decreases computation cost, and training time, and increases prediction speed. Furthermore, reducing the number of features will result in fewer electrodes, decreasing the risk of damaging the brain. Comparing the proposed algorithm to other relevant algorithms reported in the literature, the proposed algorithm has the highest average accuracy of ” ” " ~ " ” ” 98%. Also, SBT-RF acquired higher classification accuracy on four different benchmark datasets; chronic Kidney Disease (CKD), Lung Cancer (LC), Parkinson’s Disease (PD), and Stroke Prediction (SP), and it outperformed other modern techniques by 1% at least. |