On the Use of t-Distributed Stochastic Neighbor Embedding for Data Visualization and Classification of Individuals with Parkinson’s Disease

Table 4

Grand average confusion matrix of SVM for each DR method. The bold diagonal cells show the normalized (0–1) percentage of correct classifications by the SVM.

DR method			Target class
			Leave-one-out cross validation			Test set
			SH	SPD	SDBS	SH	SPD	SDBS
PCA	Predicted class	SH	0.70	0.11	0.07	0.60	0.08	0.10
		SPD	0.24	0.78	0.20	0.31	0.79	0.26
		SDBS	0.05	0.11	0.72	0.09	0.13	0.64
Sammon’s	Predicted class	SH	0.79	0.08	0.05	0.78	0.11	0.08
		SPD	0.15	0.79	0.18	0.17	0.76	0.24
		SDBS	0.05	0.13	0.77	0.06	0.13	0.68
t-SNE	Predicted class	SH	0.98	0.01	0.01	0.77	0.08	0.07
		SPD	0.02	0.98	0.03	0.14	0.78	0.19
		SDBS	0.01	0.01	0.95	0.09	0.14	0.74