Abstract
Most methods for time series classification that attain state-of-the-art accuracy have high computational complexity, requiring significant training time even for smaller datasets, and are intractable for larger datasets. Additionally, many existing methods focus on a single type of feature such as shape or frequency. Building on the recent success of convolutional neural networks for time series classification, we show that simple linear classifiers using random convolutional kernels achieve state-of-the-art accuracy with a fraction of the computational expense of existing methods. Using this method, it is possible to train and test a classifier on all 85 ‘bake off’ datasets in the UCR archive in \(<\,2\,\hbox {h}\), and it is possible to train a classifier on a large dataset of more than one million time series in approximately 1 h.
Similar content being viewed by others
References
Bagnall A, Lines J, Bostrom A, Large J, Keogh E (2017) The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances. Data Min Knowl Disc 31(3):606–660
Bagnall A, Lines J, Vickers W, Keogh E (2019) The UEA & UCR time series classification repository. http://www.timeseriesclassification.com
Bai S, Kolter JZ, Koltun V (2018) An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. ar**v:1803.01271
Benavoli A, Corani G, Mangili F (2016) Should we really use post-hoc tests based on mean-ranks? J Mach Learn Res 17(5):1–10
Bengio Y, Courville A, Vincent P (2013) Representation learning: a review and new perspectives. IEEE Trans Pattern Anal Mach Intell 35(8):1798–1828
Bostrom A, Bagnall A (2015) Binary shapelet transform for multiclass time series classification. In: Madria S, Hara T (eds) Big data analytics and knowledge discovery. Springer, Cham, pp 257–269
Bottou L, Curtis FE, Nocedal J (2018) Optimization methods for large-scale machine learning. SIAM Rev 60(2):223–311
Boureau YL, Ponce J, LeCun Y (2010) A theoretical analysis of feature pooling in visual recognition. In: Fürnkranz J, Joachims T (eds) Proceedings of the 27th international conference on machine learning, Omnipress, USA, pp 111–118
Cox D, Pinto N (2011) Beyond simple features: a large-scale feature search approach to unconstrained face recognition. Face Gesture 2011:8–15
Cui Z, Chen W, Chen Y (2016) Multi-scale convolutional neural networks for time series classification. ar**v:1603.06995
Dau HA, Bagnall A, Kamgar K, Yeh CCM, Zhu Y, Gharghabi S, Ratanamahatana CA, Keogh E (2019) The UCR time series archive. J Autom Sinica 6(6):1293–1305
Dau HA, Keogh E, Kamgar K et al (2018) UCR time series classification archive (briefing document). https://www.cs.ucr.edu/~eamonn/time_series_data_2018/
Demšar J (2006) Statistical comparisons of classifiers over multiple data sets. J Mach Learn Res 7:1–30
Dongarra J, Gates M, Haidar A, Kurzak J, Luszczek P, Tomov S, Yamazaki I (2018) The singular value decomposition: anatomy of optimizing an algorithm for extreme scale. SIAM Rev 60(4):808–865
Farahmand A, Pourazarm S, Nikovski D (2017) Random projection filter bank for time series data. In: Guyon I, Luxburg UV, Bengio S, Wallach H, Fergus R, Vishwanathan S, Garnett R (eds) Advances in neural information processing systems, vol 30. MIT Press, Cambridge, pp 6562–6572
Franceschi J, Dieuleveut A, Jaggi M (2019) Unsupervised scalable representation learning for multivariate time series. In: Seventh international conference on learning representations, learning from limited labeled data workshop
García S, Herrera F (2008) An extension on “statistical comparisons of classifiers over multiple data sets” for all pairwise comparisons. J Mach Learn Res 9:2677–2694
Goodfellow I, Bengio Y, Courville A (2016) Deep learning. MIT Press, Cambridge
Hills J, Lines J, Baranauskas E, Mapp J, Bagnall A (2014) Classification of time series by shapelet transformation. Data Min Knowl Disc 28(4):851–881
Ismail Fawaz H, Forestier G, Weber J, Idoumghar L, Muller P (2019a) Deep learning for time series classification: a review. Data Min Knowl Disc 33(4):917–963
Ismail Fawaz H, Forestier G, Weber J, Idoumghar L, Muller P (2019b) Deep neural network ensembles for time series classification. In: International joint conference on neural networks, pp 1–6
Ismail Fawaz H, Lucas B, Forestier G, Pelletier C, Schmidt DF, Weber J, Webb GI, Idoumghar L, Muller P, Petitjean F (2019c) InceptionTime: finding AlexNet for time series classification. ar**v:1909.04939
Jarrett K, Kavukcuoglu K, Ranzato M, LeCun Y (2009) What is the best multi-stage architecture for object recognition? In: 2009 IEEE 12th international conference on computer vision, pp 2146–2153
Jimenez A, Raj B (2019) Time signal classification using random convolutional features. In: 2019 IEEE international conference on acoustics, speech and signal processing
Karlsson I, Papapetrou P, Boström H (2016) Generalized random shapelet forests. Data Min Knowl Disc 30(5):1053–1085
Kingma DP, Ba JL (2015) Adam: a method for stochastic optimization. In: Third international conference on learning representations. ar**v:1412.6980
Krizhevsky A, Sutskever I, Hinton G (2012) Imagenet classification with deep convolutional neural networks. In: Pereira F, Burges CJC, Bottou L, Weinberger KQ (eds) Advances in neural information processing systems, vol 25. Curran Associates Inc, Red Hook, pp 1097–1105
Lam SK, Pitrou A, Seibert S (2015) Numba: a LLVM-based python JIT compiler. In: Proceedings of the second workshop on the LLVM compiler infrastructure in HPC, pp 1–6
Le Nguyen T, Gsponer S, Ilie I, O’Reillly M, Ifrim G (2019) Interpretable time series classification using linear models and multi-resolution multi-domain symbolic representations. Data Min Knowl Disc 33(4):1183–1222
Lin M, Chen Q, Yan S (2014) Network in network. In: Second international conference on learning representations. ar**v:1312.4400
Lines J, Taylor S, Bagnall A (2018) Time series classification with HIVE-COTE: the hierarchical vote collective of transformation-based ensembles. ACM Trans Knowl Discov Data 12(5):52:1–52:35
Lubba CH, Sethi SS, Knaute P, Schultz SR, Fulcher BD, Jones NS (2019) catch22: CAnonical Time-series CHaracteristics. Data Min Knowl Disc 33(6):1821–1852
Lucas B, Shifaz A, Pelletier C, O’Neill L, Zaidi N, Goethals B, Petitjean F, Webb GI (2019) Proximity forest: an effective and scalable distance-based classifier for time series. Data Min Knowl Disc 33(3):607–635
Middlehurst M, Vickers W, Bagnall A (2019) Scalable dictionary classifiers for time series classification. In: Yin H, Camacho D, Tino P, Tallón-Ballesteros AJ, Menezes R, Allmendinger R (eds) Intelligent data engineering and automated learning. Springer, Cham, pp 11–19
Morrow A, Shankar V, Petersohn D, Joseph A, Recht B, Yosef N (2016) Convolutional kitchen sinks for transcription factor binding site prediction. In: NIPS workshop on machine learning in computational biology
Oquab M, Bottou L, Laptev I, Sivic J (2015) Is object localization for free? Weakly-supervised learning with convolutional neural networks. In: 2015 IEEE conference on computer vision and pattern recognition, pp 685–694
Paszke A, Gross S, Chintala S, Chanan G, Yang E, DeVito Z, Lin Z, Desmaison A, Antiga L, Lerer A (2017) Automatic differentiation in PyTorch. In: NIPS autodiff workshop
Pedregosa F, Varoquaux G, Gramfort A et al (2011) Scikit-learn: machine learning in python. J Mach Learn Res 12:2825–2830
Petitjean F, Inglada J, Gancarski P (2012) Satellite image time series analysis under time war**. IEEE Trans Geosci Remote Sens 50(8):3081–3095
Pinto N, Doukhan D, DiCarlo JJ, Cox DD (2009) A high-throughput screening approach to discovering good forms of biologically inspired visual representation. PLoS Comput Biol 5(11):1–12
Rahimi A, Recht B (2008) Random features for large-scale kernel machines. In: Platt JC, Koller D, Singer Y, Roweis ST (eds) Advances in neural information processing systems, vol 20. Curran Associates Inc, Red Hook, pp 1177–1184
Rahimi A, Recht B (2009) Weighted sums of random kitchen sinks: replacing minimization with randomization in learning. In: Koller D, Schuurmans D, Bengio Y, Bottou L (eds) Advances in neural information processing systems, vol 21. MIT Press, Cambridge, pp 1313–1320
Raza A, Kramer S (2019) Accelerating pattern-based time series classification: a linear time and space string mining approach. Knowl Inf Syst 62:1113–1141
Renard X, Rifqi M, Erray W, Detyniecki M (2015) Random-shapelet: an algorithm for fast shapelet discovery. In: IEEE international conference on data science and advanced analytics, pp 1–10
Rifkin RM, Lippert RA (2007) Notes on regularized least squares. Technical report, MIT
Saxe A, Koh PW, Chen Z, Bhand M, Suresh B, Ng A (2011) On random weights and unsupervised feature learning. In: Getoor L, Scheffer T (eds) Proceedings of the 28th international conference on machine learning, Omnipress, USA, pp 1089–1096
Schäfer P (2015) The BOSS is concerned with time series classification in the presence of noise. Data Min Knowl Disc 29(6):1505–1530
Schäfer P, Leser U (2017) Fast and accurate time series classification with WEASEL. In: Proceedings of the 2017 ACM conference on information and knowledge management, pp 637–646
Shifaz A, Pelletier C, Petitjean F, Webb GI (2020) TS-CHIEF: a scalable and accurate forest algorithm for time series classification. Data Min Knowl Discov 34:742–775
Wang Z, Yan W, Oates T (2017) Time series classification from scratch with deep neural networks: a strong baseline. In: 2017 international joint conference on neural networks, pp 1578–1585
Wistuba M, Grabocka J, Schmidt-Thieme L (2015) Ultra-fast shapelets for time series classification. ar**v:1503.05018
Yosinski J, Clune J, Bengio Y, Lipson H (2014) How transferable are features in deep neural networks? In: Ghahramani Z, Welling M, Cortes C, Lawrence ND, Weinberger KQ (eds) Advances in neural information processing systems, vol 27. MIT Press, Cambridge, pp 3320–3328
Yu F, Koltun V (2016) Multi-scale context aggregation by dilated convolutions. In: Fourth international conference on learning representations. ar**v:1511.07122
Zeiler MD, Fergus R (2014) Visualizing and understanding convolutional networks. In: Fleet D, Pajdla T, Schiele B, Tuytelaars T (eds) European conference on computer vision. Springer, Cham, pp 818–833
Acknowledgements
This material is based upon work supported by an Australian Government Research Training Program Scholarship; the Air Force Office of Scientific Research, Asian Office of Aerospace Research and Development (AOARD) under award number FA2386-18-1-4030; and the Australian Research Council under awards DE170100037 and DP190100017. The authors would like to thank Professor Eamonn Keogh and all the people who have contributed to the UCR time series classification archive. Figures showing the ranking of different classifiers and variants of Rocket were produced using code from Ismail Fawaz et al. (2019a).
Author information
Authors and Affiliations
Corresponding author
Additional information
Responsible editor: Aristides Gionis, Carlotta Domeniconi, Eyke Hüllermeier, Ira Assent.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendices
Relative accuracy
1.1 ‘Bake Off’ datasets
See Fig. 13.
1.2 Additional 2018 datasets
See Fig. 14.
‘Development’ and ‘Holdout’ datasets
Additional plots for the sensitivity analysis
See Figs. 17, 18, 19, 20, 21, 22, 23 and 24.
Resamples
We also evaluate Rocket on 10 resamples of both the ‘bake off’ and additional 2018 datasets, using the same first 10 resamples (not including the original training/test split) as in Bagnall et al. (2017). Figure 25 shows the mean rank of Rocket versus HIVE-COTE, TS-CHIEF, Shapelet Transform, Proximity Forest and BOSS on the resampled ‘bake off’ datasets. Figure 27 shows the relative accuracy of Rocket and each of the other methods on the resampled ‘bake off’ datasets. The results for HIVE-COTE, Shapelet Transform, and BOSS are taken from Bagnall et al. (2019). Figures 26 and 28 show the mean rank and relative accuracy of Rocket versus Proximity Forest and TS-CHIEF for 10 resamples of the additional 2018 datasets (published results are not available for other methods for these resamples).
The results for the resamples and the original training/test splits are very similar for both the ‘bake off’ and additional 2018 datasets. In fact, while HIVE-COTE ranks ahead of Rocket, Rocket appears to be ‘stronger’ against both HIVE-COTE and TS-CHIEF on the resamples of the ‘bake off’ datasets than on the original training/test splits. For the resampled ‘bake off’ datasets, Rocket is ahead of HIVE-COTE in terms of win/draw/loss (47/2/36), as it is for the original training/test split (45/7/33). These results confirm that the results for the original training/test split are sound, and representative of the expected performance of Rocket relative to the other methods included in the comparison.
Other methods
We also compare Rocket against four recently-proposed scalable methods for time series classification (see Sect. 2.2), namely, MrSEQL, cBOSS, MiSTiCl, and catch22. We have run each of these four methods on the ‘bake off’ datasets, the additional 2018 datasets, and the scalability experiments in terms of both training set size and time series length. We have run each method with its recommended settings per the relevant papers, and using the same experimental conditions as for Rocket in each case. We have used the Python wrapper for MrSEQL (https://github.com/alan-turing-institute/sktime), the Java version of cBOSS (https://github.com/uea-machine-learning/tsml), the Java version of MiSTiCl (https://github.com/atifraza/MiSTiCl), and the Python wrapper for catch22 (https://github.com/chlubba/catch22).
1.1 ‘Bake Off’ and additional 2018 datasets
Figures 29 and 32 show the mean rank and relative accuracy of Rocket versus MrSEQL, cBOSS, MiSTiCl, and catch 22 for the 85 ‘bake off’ datasets. Figures 30 and 33 show the same for the additional 2018 datasets. Figure 31 shows total compute time.
The results show that Rocket is significantly more accurate and, with one exception, more scalable than these methods. Rocket is considerably ahead of the most accurate of these methods, MrSEQL, in terms of win/draw/loss (54/8/23) on the ‘bake off’ datasets, and Rocket is approximately an order of magnitude faster in terms of total compute time than MrSEQL, cBOSS, and MiSTiCl. While catch22 is very fast, it is the least accurate method.
As for the ‘bake off’ datasets, Rocket is significantly more accurate than MrSEQL, cBOSS, MiSTiCl, or catch22 on the additional 2018 datasets and, with the exception of catch22, considerably faster. Again, Rocket is substantially ahead of the most accurate of these methods in terms of win/draw/loss (32/0/11). Rocket is 4 times faster than cBOSS, 16 times faster than MrSEQL, and almost 22 times faster than MiSTiCl on these datasets. Again, catch22 is the fastest but least accurate method.
Note that while catch22 was originally used in conjunction with a single decision tree, we found that this produced very low accuracy and, of several ‘off the shelf’ classifiers, random forest produced the highest accuracy. Accordingly, we have used catch22 in conjunction with a random forest classifier. MiSTiCl would not run on the ElectricDevices dataset in its published configuration. For this dataset we used MiSTiCl in conjunction with AdaBoost, rather than the default extremely randomised trees. MiSTiCl would not run at all on the Chinatown dataset, so it has been ranked behind the other methods for this dataset, and this dataset has been removed from the relevant plot in Fig. 33.
1.2 Scalability
1.2.1 Training set size
Figure 34 shows accuracy and training time versus training set size for Rocket—with 10,000 (default), 1000 and 100 kernels—and the other four methods for the Satellite Image Time Series Dataset. Figure 34 shows that MrSEQL, cBOSS, and MiSTiCl are all fundamentally less scalable than Rocket in terms of training set size. By approximately 32,000 training examples, MrSEQL is approximately 75 times slower than Rocket, MiSTiCl is approximately 200 times slower than Rocket, and cBOSS is more than 300 times slower than Rocket. Additionally, all four methods are noticeably less accurate than Rocket for the same training set size. We note that there appears to be a problem with MrSEQL with more than approximately 8000 training examples. While slower than catch22 with its default settings (i.e., 10,000 kernels), in contexts where this speed difference is important, restricted to 100 kernels Rocket is an order of magnitude faster than catch22 and still significantly more accurate (see Fig. 34).
1.2.2 Time series length
Figure 35 shows training time versus time series length for Rocket versus the other four methods for the InlineSkate dataset. Figure 35 shows that, in practice, the scalability of Rocket, cBOSS, and catch22 in terms of time series length appears to be similar—that is, approximately linear in time series length. Both MrSEQL and MiSTiCl are less scalable. MrSEQL, cBOSS, and MiSTiCl are all slower than Rocket for a given time series length.
Results for ‘Bake Off’ datasets
See Table 1.
Results for additional 2018 datasets
See Table 2.
Rights and permissions
About this article
Cite this article
Dempster, A., Petitjean, F. & Webb, G.I. ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels. Data Min Knowl Disc 34, 1454–1495 (2020). https://doi.org/10.1007/s10618-020-00701-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10618-020-00701-z