Predicting the Direction Movement of Financial Time Series Using Artificial Neural Network and Support Vector Machine

Table 11

Tuning parameter selection for the SVM model using 10-fold cross-validation.

Index type	Kernel	Accuracy	F-score	Tuning parameters
KSE-100	Linear	0.8519	0.8395	C = 964.7736
	Radial basis	0.7688	0.7564	C = 137.20,
	Polynomial	0.8438	0.8344	C = 314.52, scale = 0.5554, degree = 2
Nikkei 225	Linear	0.8022	0.7912	C = 638.0629
	RBF	0.7626	0.7354	C = 1.596,
	Polynomial	0.7828	0.7532	C = 314.52, scale = 0.5554, degree = 2
KOSPI	Linear	0.8033	0.7822	C = 4
	RBF	0.8180	0.7932	C = 150,
	Polynomial	0.8033	0.7745	C = 49.298, scale = 1.042, degree = 1
SZSE composite	Linear	0.8998	0.8790	C = 324.72
	RBF	0.8720	0.8412	C = 464.666,
	Polynomial	0.8941	0.8620	C = 110.17, scale = 0.822, degree = 2