Algorithm 3: Calculation of MLP networks’ AICg
	Input: ANN model ANN = {wi,m(k)}, embedded time series data G = {X1,…,XN}, iteration number N_Max.
	Output: MLP networks’ Akaike’s information criterion AICg
1	for i = 1: N_Max
2	for n = 1: N
3	${\hat{\mathrm{y}}}_{\mathrm{n}}=\mathrm{ANN}\left({\mathrm{X}}_{\mathrm{n}}\right)=\mathrm{g}\left(p^{\left(\mathrm{K}+1\right)}\right)$
4	$\frac{\partial \mathrm{F}}{\partial {\mathrm{w}}_{\mathrm{j},\mathrm{m}}^{\left(\mathrm{k}\right)}}=\left({{\displaystyle \prod }}_{\mathrm{n}=\mathrm{k}}^{\mathrm{K}}{w}^{\left(\mathrm{n}+1\right)}{O}^{\left(\mathrm{n}\right)}\right)Q^{\left(\mathrm{k}\right)}\left(\mathrm{j},\mathrm{m}\right)U^{\left(\mathrm{k}-1\right)}$
5	$H=M^{\mathrm{T}}{(M^{\mathrm{T}}M)}^{-1}M$
6	${\mathrm{n}}_{\mathrm{g}}\approx \mathrm{trace}\left(H\right)$
7	${\mathrm{AIC}}_{\mathrm{g}}=\ln ({{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{N}}{({\mathrm{y}}_{\mathrm{n}}-{\hat{\mathrm{y}}}_{\mathrm{n}})}^2/\mathrm{N})+2{\mathrm{n}}_{\mathrm{g}}/\mathrm{N}$
8	end for
9	end for