Particle Swarm Optimization-Based Support Vector Regression for Tourist Arrivals Forecasting

Table 3

Forecast tourist arrivals obtained using ETS, ARIMA, SARIMA, GRIDSVR, PSOSVR, and FS–PSOSVR.

Case		ETS	ARIMA	SARIMA	GRIDSVR	PSOSVR	FS–PSOSVR
Japan	MAPE (%)	8.54	12.87	7.25	7.22	6.95	5.23
	RMSE	18091.23	24665.02	11878.9	14670.95	13616.38	13308.91
Hong Kong and Macao	MAPE (%)	11.07	16.28	12.10	12.39	12.24	10.65
	RMSE	21045.26	23224.72	20043.2	21168.17	21013.52	19738.13
South Korea	MAPE (%)	13.48	13.58	9.91	11.45	11.14	7.66
	RMSE	13482.26	13515.24	8681.24	11439.02	11137.59	6807.39
The United States	MAPE (%)	4.79	10.03	3.95	5.45	4.62	3.84
	RMSE	3626.19	6508.72	2786.49	3027.27	2883.04	2218.12
Total	MAPE (%)	10.64	12.08	11.24	11.21	11.14	9.76
	RMSE	100954.40	116687.03	111400.6	108457.15	107765.04	95910.99

Bold: the superior values.