Table 2.

Errors computed in the forecasts of separate phases of the fifth COVID-19 wave in Tokyo, Osaka, and Aichi with different data sets. Each experiment was conducted by excluding a single dataset (1–7) in Table 1. None demonstrates the case wherein all datasets are used, and Optimized demonstrates the best scenario. Green and red colors are the lowest and highest error values, respectively.

		Ex.1	Ex.2	Ex.3	Ex.4	Ex.5	Ex.6	Ex.7	None	Opt.a
Tokyo	Spreadb		0.30	0.26		0.41	0.53	0.43	0.44	0.23
	Peak	0.64	0.61	0.45	0.56		0.65	0.63	0.68
	Decay	0.79	1.06		0.63		0.61	0.78	0.53	0.78
	Full wave	0.63	0.65		0.46	0.52	0.62	0.61	0.55
Osaka	Spread		0.39	0.25	0.26	0.38		0.35	0.38	0.24
	Peak	0.81	0.76	0.69	0.56		0.75	0.73	0.75
	Decay	0.55	0.44	0.42	0.52	0.36	0.37			0.41
	Full wave	0.51	0.53	0.45	0.45	0.52	0.53	0.47
Aichi	Spread		0.15	0.71	0.20	0.71	0.48	0.14	0.17
	Peak		0.75	0.55	0.79	0.55	0.44	0.49	0.49
	Decay	0.95	0.85	0.96	0.65	0.85	0.92		0.60
	Full wave		0.58	0.74	0.54	0.70	0.60	0.57	0.42

^aOpt.: Optimized inputs are: 1-1, 2-4, 5-1,6-1 and 7-1 (see, Table 1).

^bSpread (July), peak (Aug.), decay (Sept.), and all wave (July–Sept.).