Multimodal Information Fusion Deep Neural Network for Large Venue Carbon Emission Assessment Model

Abstract

Large-scale events such as concerts, sports competitions, and major conferences are critical in shaping the environment and society. These events require consideration of numerous factors, including environmental parameters of the venue, such as carbon emissions, energy consumption, and temperature. Therefore, adopting sustainable practices aligned with carbon neutrality aspirations is imperative. In this context, integrating deep learning technology holds significant promise for assessing and promoting the sustainability of such events. Our model incorporates BERT (Bidirectional encoder representations from transformers) and deep neural network (DNN) modules to construct an evaluation framework. BERT, renowned for its robust natural language processing capabilities, adeptly absorbs and encodes textual knowledge into information vector representations. The DNN component in our model establishes an overarching framework for predictive modelling, providing actionable guidance and informed decision-making tools by analyzing the assembly of large-scale venues. The experimental results show that our model outperforms other state-of-the-art models, achieving at least a 1% improvement in accuracy across four datasets.

Additionally, error analysis reveals that our model reduces errors to below 10%, outperforming other models. These experimental analyses validate the superiority and practical applicability of our model. Therefore, our model offers a promising analytical approach for improving the quality and sustainability of large-scale events in the realm of carbon emissions research within event venues.