Feature Selection Using Extreme Gradient Boosting Bayesian Optimization to upgrade the Classification Performance of...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Motor imagery (MI) brain-computer interfaces (BCIs) are crucial for assisting paralyzed individuals.
• Enhancing classification accuracy in MI-BCI systems is a primary research objective.
• Effective feature selection is vital for improving MI-BCI performance.

Purpose of the Study:

• To develop a reliable MI-BCI system with superior classification accuracy.
• To investigate the efficacy of the extreme gradient Bayesian optimization (XGBO) algorithm for feature selection in MI-BCI.
• To reduce computational time and improve efficiency in MI-BCI systems.

Main Methods:

• Extracted time-, frequency-, and spatial-related MI features.
• Employed the XGBO algorithm for optimal feature selection.
• Utilized the random forest classifier for EEG signal classification.
• Validated the system on two public EEG datasets (BCI Competition III datasets IIIa and IVa).

Main Results:

• Achieved high mean accuracies of 94.44% (Dataset IIIa) and 88.72% (Dataset IVa).
• Demonstrated superior performance compared to four state-of-the-art methods, with accuracy increases of 0.87% and 0.59%.
• The XGBO algorithm significantly reduced feature dimensionality and computational time.

Conclusions:

• The proposed XGBO-based feature selection method effectively enhances MI-BCI classification accuracy.
• This approach offers a computationally efficient solution for MI-BCI systems.
• The findings contribute to advancing BCI technology for assistive applications.