Table 5. A comparison of accuracies obtained using 4 different classifiers and 3 different dimensionality reduction techniques, in addition to the accuracies obtained when PSDs are fed to the classifiers directly are shown.

All these accuracies are reported using hold-out testing technique. All classifiers and dimensionality reduction technique hyper-parameters are fine tuned using the grid searching algorithm. The highest accuracy obtained is 93%. It is obtained using SAE followed by SVM with RBF kernel.

	PCA	Kernel PCA	SAE	No dimensionality reduction (PSD is used directly)
RBF SVM	0.83	0.84	0.93	0.84
Random Forest	0.76	0.83	0.82	0.81
Logistic Regression	0.83	0.81	0.81	0.79
Neural Network	0.82	0.86	0.91	0.82