Carbon Emission Prediction in Transportation Industry Based on Hybrid Feature Selection and an IVMD-AOO-BiLSTM

To address the nonlinear and volatile characteristics of carbon emission data sequences in the transportation industry, as well as the low prediction accuracy caused by the coupling of multiple influencing factors, this study develops a carbon emission prediction model that combines hybrid feature engineering (RF-MIC), improved variational mode decomposition (IVMD), the animated oat optimization algorithm (AOO), and bidirectional long short-term memory (Bi-LSTM). First, a hybrid feature selection method Based on random forest (RF) and the maximal information coefficient (MIC) is constructed to quantify the contribution of each factor, remove redundant disturbances, and identify the key drivers of carbon emissions. Second, a multi-objective decomposition framework based on variational mode decomposition (VMD) is constructed by using the escape optimization algorithm (ESC) and Pareto optimality to adaptively optimize the number of modes K and the penalty factor α. The original carbon emission sequence is then decomposed into a series of stationary modal components, thereby mitigating its nonlinearity and volatility. Third, an AOO-based BiLSTM hyperparameter optimization theory is established, where AOO is employed to globally optimize hyperparameters such as the number of hidden layer neurons and the learning rate of BiLSTM, preventing the model from falling into local optima. Finally, prediction sub-models based on AOO-BiLSTM are constructed for each modal component, and the predicted results of all components are integrated and reconstructed to obtain the final prediction value. The proposed model is validated using carbon emission data from China’s transportation industry from 1990 to 2023. The results show that, compared with the optimal benchmark model, the root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the proposed model are reduced by 35.77%, 40.48%, and 59.52%, respectively, demonstrating its effectiveness in predicting carbon emissions in the transportation industry.