Abstract
Multi-task Learning (MTL) makes a positive difference in many fields by improving the prediction effects of correlated tasks among multiple related data sets. Some financial Multivariate Time Series (MTS) also have a high correlation, but applications like price synchronous prediction based on MTL still lack enough attention from researchers. The future values of a certain price are not only related to its own historical values, but also related to other correlated price sequences, and this is a suitable condition for applying the MTL model. This paper constructs an MTL model for synchronous learning and predicting price time series based on the Sequence-to-Sequence (Seq2Seq) model. To obtain enough data for modeling, the Weighted Soft-dtw Barycentric Averaging (wDBA) is used as the Data Augmentation (DA) method to generate more time series data for each Forex pair based on its original OHLC bid quotes. On the testing of 8 Forex pairs’ minute-level quote data from 2020 to 2022, our model, Seq2Seq with DA, outperforms baseline models including the single Long Short-term Memory (LSTM) and Seq2Seq without DA. During the experiment, to provide a comprehensive evaluation on such a long time sequence, more than 1 million minutes, we design the Chronological Randomly-sampling Walk-forward (CRSWF) Validation for a quick evaluation. As a result, when the DA degree is 125%, on the MAE, NMAE, RMSE, and NRMSE, our model respectively reduces by 95.23%, 97.24%, 94.07%, and 95.99% than LSTM, and also reduces by 7.87%, 4.84%, 6.05%, and 1.71% than the Seq2Seq without DA.
J. Li and Q. Zhao contributed equally to this work.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Bandara, K., Hewamalage, H., Liu, Y.H., Kang, Y., Bergmeir, C.: Improving the accuracy of global forecasting models using time series data augmentation. Pattern Recogn. 120, 108148 (2021)
Bellman, R.: The theory of dynamic programming. Bull. Am. Math. Soc. 60(6), 503–515 (1954)
Caruana, R.: Multitask learning. Mach. Learn. 28, 41–75 (1997)
Cho, K., et al.: Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078 (2014)
Crawshaw, M.: Multi-task learning with deep neural networks: a survey. arXiv preprint arXiv:2009.09796 (2020)
Cuturi, M., Blondel, M.: Soft-DTW: a differentiable loss function for time-series. In: International Conference on Machine Learning, pp. 894–903. PMLR (2017)
Demir, S., Mincev, K., Kok, K., Paterakis, N.G.: Data augmentation for time series regression: applying transformations, autoencoders and adversarial networks to electricity price forecasting. Appl. Energy 304, 117695 (2021)
Du, S., Li, T., Yang, Y., Horng, S.J.: Multivariate time series forecasting via attention-based encoder-decoder framework. Neurocomputing 388, 269–279 (2020)
Gao, Z.: Stock price prediction with arima and deep learning models. In: 2021 IEEE 6th International Conference on Big Data Analytics (ICBDA), pp. 61–68. IEEE (2021)
Katsiampa, P., Corbet, S., Lucey, B.: Volatility spillover effects in leading cryptocurrencies: a BEKK-MGARCH analysis. Financ. Res. Lett. 29, 68–74 (2019)
Lee, S.W., Kim, H.Y.: Stock market forecasting with super-high dimensional time-series data using ConvLSTM, trend sampling, and specialized data augmentation. Expert Syst. Appl. 161, 113704 (2020)
Luong, M.T., Le, Q.V., Sutskever, I., Vinyals, O., Kaiser, L.: Multi-task sequence to sequence learning. arXiv preprint arXiv:1511.06114 (2015)
Petitjean, F., Forestier, G., Webb, G.I., Nicholson, A.E., Chen, Y., Keogh, E.: Dynamic time warping averaging of time series allows faster and more accurate classification. In: 2014 IEEE International Conference on Data Mining, pp. 470–479. IEEE (2014)
Petitjean, F., Ketterlin, A., Gançarski, P.: A global averaging method for dynamic time warping, with applications to clustering. Pattern Recogn. 44(3), 678–693 (2011)
Seto, S., Zhang, W., Zhou, Y.: Multivariate time series classification using dynamic time warping template selection for human activity recognition. In: 2015 IEEE Symposium Series on Computational Intelligence, pp. 1399–1406. IEEE (2015)
Sutskever, I., Vinyals, O., Le, Q.V.: Sequence to sequence learning with neural networks. In: Advances in Neural Information Processing Systems, vol. 27 (2014)
Wang, P., Wang, P.: Assessment on estimations of currency basket weights-with coefficient correction for common factor dominance. Int. J. Financ. Econ. 27(1), 1401–1418 (2022)
Wang, X., Cai, Z., Luo, Y., Wen, Z., Ying, S.: Long time series deep forecasting with multiscale feature extraction and seq2seq attention mechanism. Neural Process. Lett. 54(4), 3443–3466 (2022)
Wen, Q., et al.: Time series data augmentation for deep learning: a survey. arXiv preprint arXiv:2002.12478 (2020)
ACKNOWLEDGMENT
We deeply appreciate the funding from Grant no. 2021GH10, Grant no: 2020GH10, and Grant no: EF003/FST-FSJ/2019/GSTIC by Guangzhou Development Zone Science and Technology; Grant no. 0032/2022/A and 0091/2020/A2, by Macau FDCT; Grant no. MYRG-GRG2022 and Collaborative Research Grant (MYRG-CRG) - CRG2021-00002-ICI, by University of Macau.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Li, J., Zhao, Q., Fong, S., Yen, J. (2023). Synchronous Prediction of Asset Prices’ Multivariate Time Series Based on Multi-task Learning and Data Augmentation. In: Yang, X., et al. Advanced Data Mining and Applications. ADMA 2023. Lecture Notes in Computer Science(), vol 14180. Springer, Cham. https://doi.org/10.1007/978-3-031-46677-9_37
Download citation
DOI: https://doi.org/10.1007/978-3-031-46677-9_37
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-46676-2
Online ISBN: 978-3-031-46677-9
eBook Packages: Computer ScienceComputer Science (R0)