Table 3.

Comparison of models.

Model type	Model name	Description	Differences with the other similar model
Linear regression	RML	Multiple Linear Regression model with multiple independent variables, assuming a linear relationship between the dependent variable and the independent variables	Multiple Linear Regression model with nonlinear terms has the advantage of allowing for a more complex relationship between the dependent variable and the independent variables, which can improve the fitting capability of the model, especially for nonlinear data
	RMLNE	Multiple Linear Regression model with nonlinear terms, allowing for a more complex relationship between the dependent variable and the independent variables. Differs from RML in the inclusion of nonlinear terms
Neural network	NNBP[X]	An artificial neural network model trained using the backpropagation algorithm with 1 times the number of input variables in the hidden layer(s). Differs from NNBP[2X] in the number of neurons in the hidden layer(s)	The main difference between these two models lies in their complexity and potential learning capability. NNBP[2X] has a higher number of neurons, which increases the model's capacity to learn more complex relationships between the input and output variables. This can lead to better fitting results and more accurate predictions.A higher number of neurons also increases the risk of overfitting, as the model may become too complex and fit the noise in the training data
	NNBP[2X]	An artificial neural network model similar to NNBP[X] but with twice as many neurons in the hidden layer(s)
Recurrent neural network	NNR[Y]	A neural network model that can process sequential data, using the first 30 time steps to make predictions. Differs from NNR[0.5Y] in the number of time steps used for prediction	The main difference between these two models lies in the amount of historical information they consider when making predictions. NNR[Y] takes into account a longer sequence of past data, which may provide more context and improve the model's ability to capture temporal patterns and trends. Using more time steps also increases the computational complexity of the model and may require more data to train effectively
	NNR[0.5Y]	A neural network model similar to NNR[Y] but uses the first 15 time steps for prediction
Random forest regression	RFR100	An ensemble learning model that combines multiple decision trees for regression prediction, using 100 decision trees. Differs from RFR200 in the number of decision trees used	The increased number of decision trees in RFR200 generally leads to a more complex model, which can capture more subtle patterns in the data and potentially result in more accurate predictions. However, this comes at the cost of increased computational complexity and a higher risk of overfitting, particularly if the dataset is small
	RFR200	An ensemble learning model similar to RFR100 but uses 200 decision trees for regression prediction