A FEDformer-Based Hybrid Framework for Anomaly Detection and Risk Forecasting in Financial Time Series
DOI:
https://doi.org/10.71222/dymz5n98Keywords:
FEDformer, anomaly detection, risk forecasting, financial time series, deep learning, Transformer, frequency domainAbstract
Financial markets are inherently volatile and prone to sudden disruptions such as market crashes, flash collapses, and liquidity crises. Accurate anomaly detection and early risk forecasting in financial time series are therefore crucial for preventing systemic instability and supporting informed investment decisions. Traditional deep learning models, such as LSTM and GRU, often fail to capture long-term dependencies and complex periodic patterns in highly non-stationary financial data. To address this limitation, this study proposes a FEDformer-Based Hybrid Framework for Anomaly Detection and Risk Forecasting in Financial Time Series, which integrates the Frequency Enhanced Decomposed Transformer (FEDformer) with a residual-based anomaly detector and a risk forecasting head. The FEDformer module models temporal dynamics in both time and frequency domains, decomposing signals into trend and seasonal components for improved interpretability. The residual-based detector identifies abnormal fluctuations by analyzing prediction errors, while the risk head predicts potential financial distress using learned latent embeddings. Experiments conducted on the S&P 500, NASDAQ Composite, and Brent Crude Oil datasets (2000-2024) demonstrate the superiority of the proposed model over benchmark methods, achieving an 15.7% reduction in RMSE and a 11.5% improvement in F1-score for anomaly detection. These results confirm the model's effectiveness in capturing financial volatility, enabling reliable early-warning systems for market crash prediction and risk management.
References
1. Z. Li, J. Han, and Y. Song, "On the forecasting of highfrequency financial time series based on ARIMA model improved by deep learning," Journal of Forecasting, vol. 39, no. 7, pp. 1081-1097, 2020. doi: 10.1002/for.2677
2. Z. Fan, R. Liang, and Y. Hu, "A FEDformer-Based Hybrid Framework for Anomaly Detection and Risk Forecasting in Financial Time Series," arXiv preprint arXiv:2511.12951, 2025.
3. G. Benrhmach, K. Namir, A. Namir, and J. Bouyaghroumni, "Nonlinear autoregressive neural network and extended Kalman filters for prediction of financial time series," Journal of Applied Mathematics, vol. 2020, no. 1, p. 5057801, 2020.
4. V. Söderström, and K. Knudsen, "Interpretable Outlier Detection in Financial Data: Implementation of Isolation Forest and Model-Specific Feature Importance," 2022.
5. J. Ma, and S. Perkins, "Time-series novelty detection using one-class support vector machines," In Proceedings of the International Joint Conference on Neural Networks, 2003., July, 2003, pp. 1741-1745. doi: 10.1109/ijcnn.2003.1223670
6. M. Pirani, P. Thakkar, P. Jivrani, M. H. Bohara, and D. Garg, "A comparative analysis of ARIMA, GRU, LSTM and BiLSTM on financial time series forecasting," In 2022 IEEE International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE), April, 2022, pp. 1-6. doi: 10.1109/icdcece53908.2022.9793213
7. H. Widiputra, A. Mailangkay, and E. Gautama, "Multivariate CNNLSTM model for multiple parallel financial timeseries prediction," Complexity, vol. 2021, no. 1, p. 9903518, 2021. doi: 10.1155/2021/9903518
8. T. Zhou, Z. Ma, Q. Wen, X. Wang, L. Sun, and R. Jin, "Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting," In International conference on machine learning, June, 2022, pp. 27268-27286.
9. Z. Wang, M. K. Dahouda, H. Hwang, and I. Joe, "Explanatory LSTM-AE-Based Anomaly Detection for Time Series Data in Marine Transportation," IEEE Access, 2025.
10. G. Karlsson, "Detecting Anomalies in Imbalanced Financial Data with a Transformer Autoencoder," 2024.
