BTDE-Net: A Bayesian-Transformer Hybrid Model for Predicting Digital Economy Development

Abstract

As the digital economy rapidly evolves, predicting its future trajectory faces challenges due to data complexity and uncertainty. Existing prediction methods exhibit limitations in accuracy and generalization when handling high-dimensional, multi-source heterogeneous data. To address these shortcomings, this paper proposes a Bayesian network-based model for forecasting digital economy development (BTDE-Net). The model leverages the causal inference and uncertainty quantification capabilities of Bayesian Neural Networks (BNN) and employs the Transformer architecture to capture complex dynamic features in time series data. Additionally, the Differential Evolution (DE) algorithm is used for hyperparameter optimization, enhancing the model's stability and predictive performance. Experimental results demonstrate that BTDE-Net outperforms mainstream models on the IMF WEO and World Bank Open Data datasets, achieving MAE values of 1.4 and 1.45, RMSE values of 2.05 and 2.1, and R² values of 0.94 and 0.93, respectively. This research significantly improves the accuracy and robustness of digital economy forecasting, providing a more reliable decision-making tool for policymakers.