Transformer-based Multimodal Data Fusion for Optimized Carbon Emission Prediction Model

Carbon emission prediction is a key task in addressing climate change and achieving sustainable development. To improve the accuracy of carbon emission prediction, this paper proposes a Transformer-based multimodal data fusion optimization method. The approach integrates remote sensing data, meteorological data, and land use data, effectively combining multimodal information to overcome the limitations of traditional single-source models. In terms of model architecture, the Transformer-based self-attention mechanism efficiently handles the relationships between different data sources, enhancing the model’s predictive ability in complex environments. Experimental results show that the improved model exhibits high accuracy and robustness across multiple geographic regions and varying environmental conditions, outperforming traditional regression models and other deep learning models, especially in the carbon emission prediction task.