A Novel Multilevel-SVD Method to Improve Multistep Ahead Forecasting in Traffic Accidents Domain

Table 2

Forecasting results for I-G1.

h (week)				(%)			(%)
	MSVD + AR	SWT + AR	SWT + ANN	MSVD + AR	SWT + AR	SWT + ANN	MSVD + AR	SWT + AR	SWT + ANN
1	0.0021	0.34	0.22	99.9	98.0	98.8	99.9	99.0	99.4
2	0.0030	0.29	0.19	99.9	98.3	98.9	99.9	99.2	99.5
3	0.0007	0.44	0.29	99.9	97.6	98.4	99.9	98.8	99.2
4	0.0030	0.44	0.60	99.9	97.6	96.7	99.9	98.8	98.3
5	0.0003	0.56	1.49	99.9	96.8	91.7	99.9	98.4	95.8
6	0.0024	0.51	3.10	99.9	97.1	82.9	99.9	98.6	91.4
7	0.0041	0.63	5.19	99.9	96.5	71.7	99.9	98.3	85.8
8	0.0125	0.61	7.07	99.9	96.6	63.3	99.9	98.3	81.6
9	0.0366	5.6	—	99.8	67.8	—	99.9	83.9	—
10	0.0990	4.8	—	99.4	71.6	—	99.7	85.9	—
11	0.2512	6.5	—	98.5	64.2	—	99.3	82.1	—
12	0.6009	5.9	—	96.6	66.6	—	98.3	83.3	—
13	1.3685	10.13	—	92.2	43.7	—	96.1	71.8	—
14	2.9872	—	—	83.3	—	—	91.6	—	—
Min	0.0003	0.29	0.19	83.3	43.7	63.3	91.6	71.8	81.6
Max	2.9872	10.1	7.07	99.9	98.3	98.9	99.9	99.2	99.5
Mean 1–8	0.003	0.48	2.3	99.9	97.3	87.7	99.9	98.7	93.9
Mean 1–13	0.1834	2.8	—	98.9	84.1	—	98.9	92.0	—
Mean 1–14	0.3837	—	—	97.8	—	—	98.9	—	—