Fig. 8: The box-line scatter violin plots of forecasting error. | Humanities and Social Sciences Communications

Fig. 8: The box-line scatter violin plots of forecasting error.

From: Paying attention to distortion: improving the accuracy of multistep-ahead carbon price forecasting with shape and temporal criteria

Fig. 8

This figure visualizes the distribution of forecasting errors for both the proposed model and the comparison model across multiple experiments. The scatter plots display the actual absolute error values, while the box-and-whisker plots and violin plots offer a comprehensive depiction of the distribution of these absolute error values. a Absolute error of the 3-step-ahead forecasting of dataset 1 yielded by the MLP (MSE), CNN (MSE), LSTM (MSE), LSTM (DILATE), SVMD-LSTM (MSE), and the proposed model; b absolute error of the 3-step-ahead forecasting of dataset 2 yielded by the MLP (MSE), CNN (MSE), LSTM (MSE), LSTM (DILATE), SVMD-LSTM (MSE), and the proposed model; c absolute error of the 4-step-ahead forecasting of dataset 1 yielded by the MLP (MSE), CNN (MSE), LSTM (MSE), LSTM (DILATE), SVMD-LSTM (MSE), and the proposed model; d absolute error of the 4-step-ahead forecasting of dataset 2 yielded by the MLP (MSE), CNN (MSE), LSTM (MSE), LSTM (DILATE), SVMD-LSTM (MSE), and the proposed model; e absolute error of the 5-step-ahead forecasting of dataset 1 yielded by the MLP (MSE), CNN (MSE), LSTM (MSE), LSTM (DILATE), SVMD-LSTM (MSE), and the proposed model; f absolute error of the 5-step-ahead forecasting of dataset 2 yielded by the MLP (MSE), CNN (MSE), LSTM (MSE), LSTM (DILATE), SVMD-LSTM (MSE), and the proposed model.

Back to article page