Synthetic Minority Oversampling for Imbalanced Time Series Classification Based on Path Signature
Source abstract
Abstract
Imbalanced class distributions hinder time series classifiers by underrepresenting rare yet important events. We introduce Path Signature Synthetic Time-series Oversampling (PSSTO), a structure-preserving oversampling method that operates in path signature space to synthesize informative minority samples while pruning low-quality ones. Across 12 public datasets, PSSTO with a random forest improves classification over conventional resampling approaches on average. Pairwise Wilcoxon signed-rank tests against these approaches indicate statistically significant gains. Compared with time series-specific oversamplers, PSSTO with random forest attains the best averages on F1, G-mean, and AUC compared to the strongest alternative. These results show that structure-preserving oversampling in signature space is an effective and broadly applicable remedy for imbalanced time-series classification.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Time Series Analysis and Forecasting · Imbalanced Data Classification Techniques · Machine Learning in Healthcare
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
Division of Information and Computing Technology, College of Science and Engineering, Hamad Bin Khalifa University, Doha P.O. Box 34110, Qatar
Qatar Computing Research Institute, Hamad Bin Khalifa University, Doha P.O. Box 34110, Qatar
References (49)
Show source-provided references
1. Schubach, M., Re, M., Robinson, P.N., and Valentini, G. (2017). Imbalance-Aware Machine Learning for Predicting Rare and Common Disease-Associated Non-Coding Variants. Sci. Rep., 7.10.1038/s41598-017-03011-5 ↗
2. Imbalanced classification of mental workload using a cost-sensitive majority weighted minority oversampling strategy10.1007/s10111-017-0447-x ↗
3. A new method for long-term temperature compensation of structural health monitoring by ultrasonic guided wave10.1016/j.measurement.2025.117310 ↗
4. Robust odor detection in electronic nose using transfer-learning powered Scentformer model10.1021/acssensors.5c00630 ↗
5. SMOTE: Synthetic Minority Over-sampling Technique10.1613/jair.953 ↗
6. CID: An efficient complexity-invariant distance for time series10.1007/s10618-013-0312-3 ↗
7. Cao, H., Li, X.L., Woon, Y.K., and Ng, S.K. (2011, January 11–14). SPO: Structure Preserving Oversampling for Imbalanced Time Series Classification. Proceedings of the 2011 IEEE 11th International Conference on Data Mining, Vancouver, BC, Canada.10.1109/ICDM.2011.137 ↗
8. Integrated Oversampling for Imbalanced Time Series Classification10.1109/TKDE.2013.37 ↗
9. Gong, Z., and Chen, H. (2016, January 24–28). Model-Based Oversampling for Imbalanced Sequence Classification. Proceedings of the 25th ACM International on Conference on Information and Knowledge Management, New York, NY, USA.10.1145/2983323.2983784 ↗
10. Minority oversampling for imbalanced time series classification10.1016/j.knosys.2022.108764 ↗
11. Huang, D.S., Zhang, X.P., and Huang, G.B. (2005). Borderline-SMOTE: A New Over-Sampling Method in Imbalanced Data Sets Learning. Advances in Intelligent Computing, Springer.
12. Lyons, T. (2014). Rough paths, Signatures and the modelling of functions on streams. arXiv.
13. He, H., Bai, Y., Garcia, E.A., and Li, S. (2008, January 1–8). ADASYN: Adaptive synthetic sampling approach for imbalanced learning. Proceedings of the 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), Hong Kong, China.
14. A study of the behavior of several methods for balancing machine learning training data10.1145/1007730.1007735 ↗
15. Sampling technique for noisy and borderline examples problem in imbalanced classification10.1016/j.asoc.2023.110361 ↗
16. DeepSMOTE: Fusing Deep Learning and SMOTE for Imbalanced Data10.1109/TNNLS.2021.3136503 ↗
17. A new imbalanced data oversampling method based on Bootstrap method and Wasserstein Generative Adversarial Network10.3934/mbe.2024190 ↗
18. Kotelnikov, A., Baranchuk, D., Rubachev, I., and Babenko, A. (2023, January 23–29). TabDDPM: Modelling tabular data with diffusion models. Proceedings of the 40th International Conference on Machine Learning, Honolulu, HI, USA. ICML’23.
19. Schreyer, M., Sattarov, T., Sim, A., and Wu, K. (2024, January 14–17). Imb-FinDiff: Conditional Diffusion Models for Class Imbalance Synthesis of Financial Tabular Data. Proceedings of the 5th ACM International Conference on AI in Finance, New York, NY, USA. ICAIF ’24.10.1145/3677052.3698659 ↗
20. Zhao, P., Luo, C., Qiao, B., Wang, L., Rajmohan, S., Lin, Q., and Zhang, D. (2022, January 23–29). T-SMOTE: Temporal-oriented Synthetic Minority Oversampling Technique for Imbalanced Time Series Classification. Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, Vienna, Austria.10.24963/ijcai.2022/334 ↗
21. Cranefield, S. (2013). An Effective Method for Imbalanced Time Series Classification: Hybrid Sampling. AI 2013: Advances in Artificial Intelligence, Springer International Publishing.
22. Imbalanced Data Problem in Machine Learning: A Review10.1109/ACCESS.2025.3531662 ↗
23. Integration of Paths–A Faithful Representation of Paths by Noncommutative Formal Power Series
24. Uniqueness for the signature of a path of bounded variation and the reduced path group10.4007/annals.2010.171.109 ↗
25. The UCR Time Series Archive10.1109/JAS.2019.1911747 ↗
26. Time Series FeatuRe Extraction on basis of Scalable Hypothesis tests (tsfresh—A Python package)10.1016/j.neucom.2018.03.067 ↗
27. catch22: CAnonical Time-series CHaracteristics10.1007/s10618-019-00647-x ↗
28. Experiencing SAX: A novel symbolic representation of time series10.1007/s10618-007-0064-z ↗
29. The BOSS is concerned with time series classification in the presence of noise10.1007/s10618-014-0377-7 ↗
30. Classification of time series by shapelet transformation10.1007/s10618-013-0322-1 ↗
31. Lyons, T.J., Caruana, M., and Lévy, T. (2007). Differential Equations Driven by Rough Paths, Springer.10.1007/978-3-540-71285-5 ↗
32. Levin, D., Lyons, T., and Ni, H. (2016). Learning from the past, predicting the statistics for the future, learning an evolving system. arXiv.
33. Discretely sampled signals and the rough Hoff process10.1016/j.spa.2016.02.011 ↗
34. Chevyrev, I., and Kormilitzin, A. (2016). A Primer on the Signature Method in Machine Learning. arXiv.
35. Lucchese, L., Pakkanen, M.S., and Veraart, A.E.D. (2025, January 13–19). Learning with Expected Signatures: Theory and Applications. Proceedings of the Forty-second International Conference on Machine Learning, Vancouver, BC, Canada.
36. Chang, J., and Lyons, T. (2019). Insertion algorithm for inverting the signature of a path. arXiv.
37. Fermanian, A., Chang, J., Lyons, T., and Biau, G. (2024). The Insertion Method to Invert the Signature of a Path. Recent Advances in Econometrics and Statistics, Springer Nature.10.1007/978-3-031-61853-6_29 ↗
38. Morrill, J., Fermanian, A., Kidger, P., and Lyons, T. (2021). A generalised signature method for multivariate time series feature extraction. arXiv.
39. Session details: Special issue on learning from imbalanced datasets10.1145/1007730.1007733 ↗
40. Kubát, M., and Matwin, S. (1997, January 8–12). Addressing the Curse of Imbalanced Training Sets: One-Sided Selection. Proceedings of the International Conference on Machine Learning, Nashville, TN, USA.
41. Saito, T., and Rehmsmeier, M. (2015). The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets. PLoS ONE, 10.10.1371/journal.pone.0118432 ↗
42. An introduction to ROC analysis10.1016/j.patrec.2005.10.010 ↗
43. Glazkova, A. (2020). A Comparison of Synthetic Oversampling Methods for Multi-class Text Classification. arXiv.
44. Xiong, H., and Lee, W.B. (2011). A New Over-Sampling Approach: Random-SMOTE for Learning from Imbalanced Data Sets. Knowledge Science, Engineering and Management, Springer.10.1007/978-3-642-25975-3_30 ↗
45. Improving imbalanced learning through a heuristic oversampling method based on k-means and SMOTE10.1016/j.ins.2018.06.056 ↗
46. Smote-variants: A python implementation of 85 minority oversampling techniques10.1016/j.neucom.2019.06.100 ↗
47. An empirical comparison and evaluation of minority oversampling techniques on a large number of imbalanced datasets10.1016/j.asoc.2019.105662 ↗
48. Secure, privacy-preserving and federated machine learning in medical imaging10.1038/s42256-020-0186-1 ↗
49. Torkzadehmahani, R., Kairouz, P., and Paten, B. (2019, January 16–17). DP-CGAN: Differentially private synthetic data and label generation. Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Long Beach, CA, USA.10.1109/CVPRW.2019.00018 ↗