Fig 1.
ANN architecture used in this study.
Fig 2.
Average monthly temperature of Upper East from 1990 to 2021.
Table 1.
Descriptive statistics of the mean monthly temperature of Upper East and Upper West regions, Ghana.
Fig 3.
Flowchart of the three-phase hybrid CEEMDAN-LMD-NN model integrating complete ensemble empirical mode decomposition adaptive noise (CEEMDAN), local mean decomposition (LMD), and neural network (NN) algorithm.
Fig 4.
Decomposition of original signal using CEEMDAN to IMFs.
Table 2.
Descriptive statistics of the mean monthly temperature of Upper East and Upper West regions, Ghana.
Table 3.
Sample entropy of each CEEMDAN IMF and residue signal.
Fig 5.
Density of the mean temperature for the original total period, training data, and testing data.
Table 4.
Summary statistics of the total training and testing temperature data.
Fig 6.
Decomposition of the IMF1 using LMD to PFs.
Table 5.
The 2-month, 4-month, 6-month ahead prediction performance comparison of the temperature data for different standalone prediction models.
Table 6.
The 2-step, 4-step, 6-step-ahead prediction performance comparison of the temperature data for different two-phase prediction models.
Fig 7.
A comparison of RMSE, MAPE, MAE, P-Bias, and RSR for the step-ahead forecasts generated by the different hybrid three phase prediction models.
Fig 8.
Taylor plot for CEEMDAN-LMD-ARIMA, CEEMDAN-LMD-ELM, and CEEMDAN-LMD-NN.
Table 7.
The 2-step, 4-step, 6-step-ahead prediction performance comparison of the temperature data for different three-phase prediction models.
Table 8.
The 2-step, 4-step, 6-step-ahead prediction performance comparison of the temperature data for different three-phase prediction models.