Rough set-based classification of EEG-signals to detect intraoperative awareness: Comparison of fuzzy and crisp discretization of real value attributes

Automated classification of calculated EEG parameters has been shown to be a promising method for detection of intraoperative awareness. In the present study, rough set-based methods were employed to generate classification rules. For these methods, discrete attributes are required. We compared a crisp and a fuzzy discretization of the real parameter values. Fuzzy discretization transforms one real attribute value to several discrete values. By combining the different (discrete) values of all attributes, several sub-objects were produced from a single original object. Rule generation from a training set of objects and classification of a test set provided good classification rates of approximately 90% for both crisp and fuzzy discretization. Fuzzy discretization resulted in a simpler and smaller rule set than crisp discretization. Therefore, the simplicity of the resulting classifier justifies the higher computational effort caused by fuzzy discretization.     «

Automated classification of calculated EEG parameters has been shown to be a promising method for detection of intraoperative awareness. In the present study, rough set-based methods were employed to generate classification rules. For these methods, discrete attributes are required. We compared a crisp and a fuzzy discretization of the real parameter values. Fuzzy discretization transforms one real attribute value to several discrete values. By combining the different (discrete) values of all at...     »