Carbon Emission Prediction in Industrial Zones Using IGWO-Optimized SVM and STIRPAT

Abstract

With the continuous advancement of technology, carbon emission prediction results have become increasingly reliable. However, traditional carbon emission prediction methods face limitations in data and uncertainty, requiring substantial experimental resources and data. Therefore, the study optimizes the Support Vector Machine through the Improved Grey Wolf Optimizer. It combines the extended stochastic environmental impact assessment model to provide a framework for influencing factors and introduces randomness and regression analysis. This approach improves the accuracy and applicability of the fusion model in predicting carbon emissions in industrial zones. Experimental results show that, in the Hubei Province dataset, the proposed model achieves the smallest Mean Squared Error of 0.0075 among four models. The Root Mean Squared Error values of the individual Feedforward Neural Network and Multilayer Perceptron are 0.0101 and 0.0197 higher than that of the proposed model, respectively. Compared to existing single models, such as backpropagation neural networks, the Root Mean Squared Error values of the studied model is significantly reduced by 12%. These results indicate that the proposed prediction model demonstrates excellent timeliness in carbon emission forecasting. This capability provides policy makers with a variety of policy assessment tools to help develop more effective emissions reduction policies.