Random Forest Classification for Brain Computer Interface Applications


OKUMUŞ H. , AYDEMİR Ö.

25th Signal Processing and Communications Applications Conference (SIU), Antalya, Türkiye, 15 - 18 Mayıs 2017 identifier identifier

Özet

Brain computer interface applications have big importance in becoming a bridge between the human brain and devices. The studies in this area increase every day with the use of different feature extractions and classification methods In this study, classification is done by Random Forest method using Data Set HI presented in BCI Competiton 2003, and it has been shown that combining the Fast Walsh Hadamard Transformation with Fourier and Wavelet Transform is effective in the classification accuracy. The frequently used Support Vector Machines, k-Nearest Neighbor, and Linear Separator Analysis methods are also applied in this study and the results are compared with the literature results which have used the same dataset The highest classification accuracy was obtained as 89.06% when the specified conversion methods were used together and the classification method was Random Forest, indicating that the preferred classification method and conversion methods are effective when compared to other classifiers in the literature.