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

Table 4

Multistep MIMO forecasting results, nRMSE, mNSE, and mIA for I-G3.

h; (week)	nRMSE			mNSE(%)			mIA(%)
	MSVD + AR	SWT + AR	SWT + ANN	MSVD + AR	SWT + AR	SWT + ANN	MSVD + AR	SWT + AR	SWT + ANN
1	0.0001	0.48	0.46	99.9	97.7	97.6	99.9	98.9	98.8
2	0.0020	0.40	0.41	99.9	98.0	97.9	99.9	99.0	98.9
3	0.0005	0.57	0.52	99.9	97.2	97.3	99.9	98.6	98.6
4	0.0007	0.57	1.58	99.9	97.3	91.9	99.9	98.6	95.9
5	0.0004	0.72	3.76	99.9	96.5	81.4	99.9	98.3	90.7
6	0.0015	0.63	7.03	99.9	96.9	65.6	99.9	98.4	82.8
7	0.0052	0.82	9.54	99.9	96.1	53.2	99.9	98.0	76.6
8	0.0165	0.82	9.63	99.9	96.1	52.4	99.9	98.1	76.2
9	0.0482	7.83	—	99.8	62.5	—	99.9	81.3	—
10	0.1303	6.59	—	99.4	67.7	—	99.7	83.9	—
11	0.3293	8.44	—	98.4	59.6	—	99.2	79.8	—
12	0.7864	7.96	—	96.2	61.8	—	98.1	80.9	—
13	1.7868	13.04	—	91.4	35.0	—	95.7	67.5	—
14	3.8944	—	—	81.4	—	—	90.7	—	—
Min	0.0001	0.56	0.41	81.4	35.0	52.4	90.7	67.5	76.2
Max	3.89	19.6	9.6	99.9	98.0	97.9	99.9	99.0	98.9
Mean 1–8	0.003	0.63	4.12	99.9	96.9	79.7	99.9	98.5	89.8
Mean 1–13	0.24	5.46	—	98.9	81.7	—	99.4	90.9	—
Mean 1–14	0.5	—	—	97.6	—	—	98.8	—	—