Table 1 ACC, RMSE of the predictions and model configuration of the optimum ML model.

March

[0.39; 2.97]

{0.38; 2.97}

[0.72; 2.00]

[0.27; 2.71]

{0.34; 2.63}

Model configuration:

AdaBoost (MLP); Norm; avg of (RELU ADAM, RELU LBFGS, TANH ADAM, TANH LBFGS; nproc:2–20 )

April

[0.15; 3.77]

{0.22; 3.45}

[0.49; 2.64]

[0.33; 2.77]

{0.35; 2.76}

Model configuration:

AdaBoost (MLP); Norm; PCA; RELU ADAM, avg of nproc:15 and 16

May Reg1

[~ 0.0; 3.53]

{0.19; 3.41}

[0.47; 2.3]

[0.33; 2.56]

{0.28; 2.80}

Model configuration:

AdaBoost (MLP); Std; PCA; avg of (RELU ADAM, TANH LBFGS; nprocs:10–20)

May Reg2

[0.08; 2.81]

{0.18; 2.78}

[0.32; 2.41]

[0.36; 2.42]

{0.28; 2.74}

Model configuration:

AdaBoost (MLP); Std; RELU ADAM, avg of nproc: 10–20

June

[0.38; 2.70]

{0.32; 2.76}

[0.62; 2.16]

[0.46; 2.25]

{0.44; 2.23}

Model configuration:

AdaBoost (MLP); RELU ADAM, avg of (Norm, Norm PCA, Std, Std PCA, Power, Robust; nproc: 2–14)

[corr; rmse] for the period 1999–2020; {corr; rmse} for the period 1982–2020

corr: correlation coefficient; rmse: root mean square error

Std: Standardization; Norm: Min–Max Normalization

Activation Function: RELU, TANH; Solver: ADAM, LBFGS

nproc: Number of processors; PCA: Principal component analysis

MLP: Multi layer perceptron