A novel binary data classification system based on the modified Gray-Scott model

초록

We propose a modified Gray-Scott (GS) model for binary data classification. We assign the corresponding initial values to the concentrations of reactants U and V in the GS model, and discretize them on a grid. In each simulation iteration, we retain the initial value of V for calculation to determine the fuzzy boundary so that the concentration of the reactants remains within a certain range near the boundary during diffusion. This ultimately stabilizes the diffusion process and forms a nonlinear interface or hyperplane. We conduct various computational tests to verify the efficiency and robustness of the proposed algorithm. In addition, classification problems in real-world applications, such as those involving Electroencephalogram (EEG) signals, are also considered. Some statistical metrics such as Shannon entropy (SNE), Higuchi's Hurst exponent (HHE), Kolmogorov complexity (KC), and Hurst exponents H(-10)$H(-10)$, H(0), and H(10), which are extracted from q=-10,0,10$q=-10,0,10$ by Multiple Fractal Detrending Fluctuation Analysis (MF-DFA) were used to characterize the EEG signals. The advantage of this method compared to traditional machine learning classification methods is that it can construct a nonlinear interface or hyperplane, which can significantly improve classification accuracy, recall and precision.

키워드