Deep learning-based enterprise operation forecasting and green production optimization under the BRI

The BRI (BRI) has triggered radical changes in international trade, infrastructure construction, and transnational collaboration and offered businesses opportunities and challenges never seen before. Since businesses are struggling to optimize their operations and meet the obligation of environmental sustainability and corporate social responsibility (CSR), a heterogeneous market with dynamic constraints in BRI areas would make decision−making more difficult. To overcome these obstacles, this paper will present a new deep learning−based optimization model, DWT−CNN−BiGRU with Attention and Social−Environmental Constraints that aims at predicting the operations of the enterprise and planning green production under the BRI. The architecture uses bidirectional gated recurrent units (BiGRU) to estimate time−dependent relationships, convolutional neural networks (CNN) to extract features, a discrete wavelet transform (DWT) to denoise, and an attention mechanism to weight and interpret dynamically. It is demonstrated by the experimental results that the proposed model has a higher prediction accuracy, as compared to the baseline models (LSTM, GRU, BiLSTM, CNN−LSTM and CNN−BiLSTM), on multiple real−world datasets, such as global tourism, China city air quality, and tourism−pollution datasets. RMSE, MAE, and MAPE are decreased up to 12.01, 11.49, and 12.01 respectively. It also enables the framework to incorporate green production constraints, which would enhance the performance of CSR−related activities and reduce the estimated emission of energy and carbon by 15 to 20%.