Improving Accuracy of River Flow Forecasting Using LSSVR with Gravitational Search Algorithm

Table 2

Comparison of HLGSA, MARS, M5RT, and CMLR methods: Astore catchment.

Indexes	Cross validation data sets	Input combinations	Methods
			HLGSA	MARS	M5RT	CMLR
	DS1	IC1	56.80	57.75	71.96	83.44
		IC2	44.80	53.48	56.66	59.75
		IC3	51.01	55.28	80.52	55.66
	DS2	IC1	72.58	73.83	83.09	97.06
		IC2	65.42	67.06	68.28	69.44
		IC3	66.97	68.07	101.99	68.08
	DS3	IC1	58.27	65.16	66.36	91.31
		IC2	47.38	51.59	59.57	69.20
		IC3	47.61	56.20	67.22	65.93
	DS4	IC1	46.93	57.75	61.29	74.79
		IC2	40.09	43.46	57.26	61.40
		IC3	42.57	48.14	59.20	58.08
	DS1	IC1	34.32	38.34	42.05	51.91
		IC2	27.03	32.30	33.45	26.36
		IC3	31.67	33.75	46.80	25.05
	DS2	IC1	44.24	49.12	52.21	57.67
		IC2	37.86	37.96	39.12	42.11
		IC3	38.27	38.92	53.77	40.25
	DS3	IC1	33.80	35.79	40.08	52.33
		IC2	27.14	29.23	32.47	34.33
		IC3	28.81	30.84	42.56	31.06
	DS4	IC1	32.20	37.19	38.46	46.80
		IC2	25.05	28.36	29.83	39.48
		IC3	26.36	30.22	33.74	32.70
	DS1	IC1	0.831	0.822	0.732	0.693
		IC2	0.889	0.851	0.835	0.815
		IC3	0.852	0.849	0.745	0.837
	DS2	IC1	0.826	0.820	0.729	0.685
		IC2	0.850	0.846	0.822	0.811
		IC3	0.848	0.834	0.630	0.824
	DS3	IC1	0.849	0.830	0.773	0.694
		IC2	0.901	0.880	0.851	0.831
		IC3	0.898	0.861	0.810	0.844
	DS4	IC1	0.892	0.882	0.813	0.699
		IC2	0.916	0.900	0.888	0.853
		IC3	0.906	0.893	0.874	0.882