KOSLM: A Kalman-Optimal Hybrid State-Space Memory Network for Long-Term Time Series Forecasting
Source abstract
Abstract
Long-term time series forecasting (LTSF) remains challenging, as models must capture long-range dependencies and remain robust to noise accumulation. Traditional recurrent models, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM), often suffer from instability and information degradation over extended horizons. The state-of-the-art method xLSTMTime improves memory retention through exponential gating and enhanced memory-transition rules, but it still lacks principled guidance. To address these issues, we propose the Kalman-Optimal Selective Long-Term Memory (KOSLM) model, which embeds a Kalman-optimal selective mechanism driven by the innovation signal within a structured state-space reformulation of LSTM. KOSLM dynamically regulates information propagation and forgetting to minimize state estimation uncertainty, providing both theoretical interpretability and practical efficiency. Extensive experiments across energy, finance, traffic, healthcare, and meteorology datasets show that KOSLM reduces mean squared error (MSE) by 14.3–38.9% compared with state-of-the-art methods, with larger gains at longer horizons. The model is lightweight, scalable, and achieves up to 2.5× speedup over Mamba-2. Beyond benchmarks, KOSLM is further validated on real-world Secondary Surveillance Radar (SSR) tracking under noisy and irregular sampling, demonstrating robust and generalizable long-term forecasting performance.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Traffic Prediction and Management Techniques · Forecasting Techniques and Applications · Meteorological Phenomena and Simulations
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Electronic Information and Artificial Intelligence, Shaanxi University of Science and Technology, Xi’an 710021, China
References (41)
Show source-provided references
1. Cheng, M., Liu, Z., Tao, X., Liu, Q., Zhang, J., Pan, T., Zhang, S., He, P., Zhang, X., and Wang, D. (2025). A comprehensive survey of time series forecasting: Concepts, challenges, and future directions. TechRxiv.10.36227/techrxiv.174430535.53879341/v1 ↗
2. A distance correlation-based approach to characterize the effectiveness of recurrent neural networks for time series forecasting10.1016/j.neucom.2025.129641 ↗
3. Long short-term memory10.1162/neco.1997.9.8.1735 ↗
4. Enhancing continuous integration predictions: A hybrid LSTM-GRU deep learning framework with evolved DBSO algorithm10.1007/s00607-024-01370-2 ↗
5. Pascanu, R., Mikolov, T., and Bengio, Y. (2013, January 16–21). On the Difficulty of Training Recurrent Neural Networks. Proceedings of the 30th International Conference on Machine Learning (ICML’13), Atlanta, GA, USA.
6. A New Approach to Linear Filtering and Prediction Problems10.1115/1.3662552 ↗
7. Robustness enhancement in neural networks with alpha-stable training noise10.1016/j.dsp.2024.104778 ↗
8. Krishnan, R.G., Shalit, U., and Sontag, D. (2015). Deep kalman filters. arXiv.
9. KalmanNet: Neural network aided Kalman filtering for partially known dynamics10.1109/TSP.2022.3158588 ↗
10. Dahan, Y., Revach, G., Dunik, J., and Shlezinger, N. (2024, January 14–19). Uncertainty quantification in deep learning based Kalman filters. Proceedings of the ICASSP 2024—2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Seoul, Republic of Korea.10.1109/ICASSP48485.2024.10447987 ↗
11. Gu, A., and Dao, T. (2023). Mamba: Linear-time sequence modeling with selective state spaces. arXiv.
12. Dao, T., and Gu, A. (2024). Transformers are ssms: Generalized models and efficient algorithms through structured state space duality. arXiv.
13. Katsikis, V.N. (2012). Optimal Solution to Matrix Riccati Equation—For Kalman Filter Implementation. MATLAB, IntechOpen. Chapter 4.
14. On the parameterization and initialization of diagonal state space models
15. Gu, A., Gulcehre, C., Paine, T., Hoffman, M., and Pascanu, R. (2020, January 13–18). Improving the gating mechanism of recurrent neural networks. Proceedings of the International Conference on Machine Learning, PMLR, Virtual.
16. (2024, May 12). Weather Dataset. 2024. Weather Data from the BGC Jena Climate Data Archive. Available online: https://www.hypermeteo.com/en/?gad_source=1&gad_campaignid=14712996685&gclid=EAIaIQobChMIvJ_HqOORkQMVSpCDBx3MFTosEAAYASAAEgLx7fD_BwE.
17. (2024, May 12). ETT Datasets. 2024. Energy Transformer Temperature Dataset (ETT) for Long-Term Forecasting. Available online: https://ieee-dataport.org/documents/merged-ett-dataset-time-series-forecasting.
18. (2024, May 12). Electricity Dataset. 2024. Electricity Load Diagrams Dataset from UCI Machine Learning Repository. Available online: https://ieee-dataport.org/documents/uci-dataset.
19. (2024, May 12). Exchange Rate Dataset. 2024. Daily Exchange Rate Dataset for Long-Term Time Series Forecasting, Provided with the ETT Benchmark Collection. Available online: https://github.com/juyongjiang/TimeSeriesDatasets.
20. (2024, May 12). Traffic Dataset. 2024. Traffic Flow Data from California Performance Measurement System (PeMS). Available online: https://www.smartmicro.com/traffic-sensor?utm_medium=advert&utm_source=google&utm_campaign=search-traffic&utm_content=ad&utm_term=traffic_sensor&gad_source=1&gad_campaignid=22435495111&gclid=EAIaIQobChMIlvLc8OORkQMVQJKDBx3FOSSQEAAYASAAEgITNPD_BwE.
21. (2024, May 12). Illness (ILI) Dataset. 2024. Weekly Influenza-Like Illness (ILI) Dataset for Forecasting Tasks, Available Within the ETT Benchmark Suite. Available online: https://www.researchgate.net/figure/Influenza-like-illness-ILI-data-from-the-FluID-database-Weekly-scaled-rates-of-ILI_fig1_337718309.
22. xlstmtime: Long-term time series forecasting with xlstm10.3390/ai5030071 ↗
23. Film: Frequency improved legendre memory model for long-term time series forecasting
24. Is mamba effective for time series forecasting?10.1016/j.neucom.2024.129178 ↗
25. Zhou, T., Ma, Z., Wen, Q., Wang, X., Sun, L., and Jin, R. (2022, January 17–23). Fedformer: Frequency enhanced decomposed transformer for long-term series forecasting. Proceedings of the International Conference on Machine Learning, PMLR, Baltimore, MD, USA.
26. Liu, Y., Hu, T., Zhang, H., Wu, H., Wang, S., Ma, L., and Long, M. (2023). itransformer: Inverted transformers are effective for time series forecasting. arXiv.
27. Zhang, Y., and Yan, J. (2023, January 1–5). Crossformer: Transformer Utilizing Cross-Dimension Dependency for Multivariate Time Series Forecasting. Proceedings of the Eleventh International Conference on Learning Representations (ICLR), Kigali, Rwanda.
28. Su, J., Shen, Y., and Wu, Z. (2022). DLinear: A Linear Complexity Approach for Long-Term Time Series Forecasting. arXiv.
29. Long time series of ocean wave prediction based on PatchTST model10.1016/j.oceaneng.2024.117572 ↗
30. Wang, S., Wu, H., Shi, X., Hu, T., Luo, H., Ma, L., Zhang, J.Y., and Zhou, J. (2024). Timemixer: Decomposable multiscale mixing for time series forecasting. arXiv.
31. Dao, T. (2024). Flashattention-2: Faster attention with better parallelism and work partitioning. arXiv.
32. Study of ADS-B data evaluation10.1016/S1000-9361(11)60053-8 ↗
33. Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., and Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv.10.3115/v1/D14-1179 ↗
34. Mehta, H., Gupta, A., Cutkosky, A., and Neyshabur, B. (2022). Long range language modeling via gated state spaces. arXiv.
35. Hua, W., Dai, Z., Liu, H., and Le, Q. (2022, January 23–29). Transformer quality in linear time. Proceedings of the International Conference on Machine Learning, PMLR, Honolulu, HI, USA.
36. Ha, D., Dai, A., and Le, Q.V. (2016). Hypernetworks. arXiv.
37. Poli, M. (2023). Data-dependent neural networks. arXiv.
38. Combining recurrent, convolutional, and continuous-time models with structured state space
39. Approximation of dynamical systems by continuous time recurrent neural networks10.1016/S0893-6080(05)80125-X ↗
40. Tallec, C., and Ollivier, Y. (2018). Can recurrent neural networks warp time?. arXiv.
41. Tahirovic, A., and Redzovic, A. (2025). Optimal Estimation for Continuous-Time Nonlinear Systems Using State-Dependent Riccati Equation (SDRE). arXiv.