Enhancing exchange rate prediction and risk management under uncertainty shocks: an AI-driven ensemble prediction model based on metaheuristic optimization


Sun W., Li M., Chen X. H., Wang Y.

ANNALS OF OPERATIONS RESEARCH, 2024 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Publication Date: 2024
  • Doi Number: 10.1007/s10479-024-06319-4
  • Journal Name: ANNALS OF OPERATIONS RESEARCH
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, ABI/INFORM, Applied Science & Technology Source, Business Source Elite, Business Source Premier, Computer & Applied Sciences, INSPEC, Public Affairs Index, zbMATH, Civil Engineering Abstracts
  • Azerbaijan State University of Economics (UNEC) Affiliated: No

Abstract

The foreign exchange market significantly affects international trade, making accurate exchange rate predictions essential for investors, businesses, and government organizations. In this study, we propose an AI-driven ensemble model called the multi-model dynamic prediction system (MMDPS) to enhance exchange rate forecasting accuracy, prediction direction consistency, and prediction robustness. To that end, we implemented a series of measures. First, the submodels of MMDPS apply fuzzy prediction, chaos prediction, deep learning, and linear time-series forecasting methods to enhance the information extraction capability of MMDPS. Second, we introduced an innovative hybrid wavelet denoising method to enhance the ability of MMDPS to handle noise in exchange rate data. Third, we employed a novel multi-objective metaheuristic optimization algorithm to determine the optimal combination weights, optimizing across the dimensions of predictive accuracy and directional consistency. This approach further enhances the generalization capability and predictive robustness of MMDPS. Finally, we incorporated into MMDPS multiple variables that are closely related to exchange rates and can rapidly reflect market changes, such as panic indices and commodity futures prices, to enhance the model's predictive robustness against uncertainty shocks. The predictive performance of MMDPS was validated using 15 datasets with varying frequencies. The results of our experiments demonstrate that MMDPS significantly outperforms the benchmark models in terms of prediction accuracy, consistency in prediction direction, and predictive robustness. Furthermore, MMDPS can significantly enhance investor returns and mitigate investment risks, which underscore its critical application value.