Air Pollution Prediction Using an Ensemble of Dynamic Transfer Models for Multivariate Time Series
Abstract
:1. Introduction
2. Preliminaries: ARIMA, Transfer Function, Ensemble
3. Proposed Method
3.1. Dynamic Transfer Models
3.2. Ensemble of Dynamic Transfer Models
- Step 1.
- Generate numerous candidate models for response as in (1) with and using training observations.
- Step 2.
- Choose the top-K models as base learners among the L candidate models in terms of minimum prediction error.
- Step 3.
- Generate prediction models for input variables, , using training observations.
4. Experiments
4.1. Simulation 1
4.2. Simulation 2
4.3. Application 1: Air Quality Data
4.4. Application 2: APM Data
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Barford, P.; Kline, J.; Plonka, D.; Ron, A. A signal analysis of network traffic anomalies. In Proceedings of the 2nd ACM SIGCOMM Workshop on Internet Measurment, Marseille, France, 6–8 November 2002; pp. 71–82. [Google Scholar]
- Kruegel, C.; Vigna, G. Anomaly detection of web-based attacks. In Proceedings of the 10th ACM conference on Computer and Communications Security, Washington, DC, USA, 27–30 October 2003; pp. 251–261. [Google Scholar]
- Quenouille, M.H.; Quenouille, M.H. The Analysis of Multiple Time-Series; No. 519.232 Q84; Griffin: London, UK, 1957. [Google Scholar]
- Veenstra, A.W.; Haralambides, H.E. Multivariate autoregressive models for forecasting seaborne trade flows. Transp. Res. Part E Logist. Transp. Rev. 2001, 37, 311–319. [Google Scholar] [CrossRef]
- Chevillon, G.; Hendry, D.F. Non-parametric direct multi-step estimation for forecasting economic processes. Int. J. Forecast. 2005, 21, 201–218. [Google Scholar] [CrossRef] [Green Version]
- Haldrup, N.; Nielsen, F.S.; Nielsen, M.Ø. A vector autoregressive model for electricity prices subject to long memory and regime switching. Energy Econ. 2010, 32, 1044–1058. [Google Scholar] [CrossRef] [Green Version]
- Gong, C.; Tang, P.; Wang, Y. Measuring the network connectedness of global stock markets. Phys. Stat. Mech. Its Appl. 2019, 535, 122351. [Google Scholar] [CrossRef]
- Huarng, K.H.; Yu, T.H.K.; Hsu, Y.W. A multivariate heuristic model for fuzzy time-series forecasting. IEEE Trans. Syst. Man Cybern. Part B (Cybern.) 2007, 37, 836–846. [Google Scholar] [CrossRef] [PubMed]
- Egrioglu, E.; Aladag, C.H.; Yolcu, U.; Uslu, V.R.; Basaran, M.A. A new approach based on artificial neural networks for high order multivariate fuzzy time series. Expert Syst. Appl. 2009, 36, 10589–10594. [Google Scholar] [CrossRef]
- Isufi, E.; Loukas, A.; Perraudin, N.; Leus, G. Forecasting time series with varma recursions on graphs. IEEE Trans. Signal Process. 2019, 67, 4870–4885. [Google Scholar] [CrossRef] [Green Version]
- Adhikari, R.; Agrawal, R.K. A novel weighted ensemble technique for time series forecasting. In Pacific-Asia Conference on Knowledge Discovery and Data Mining; Springer: Berlin/Heidelberg, Germany, 2012; pp. 38–49. [Google Scholar]
- Borovykh, A.; Oosterlee, C.W.; Bohté, S.M. Generalization in fully-connected neural networks for time series forecasting. J. Comput. Sci. 2019, 36, 101020. [Google Scholar] [CrossRef]
- Wang, X.; Han, M. Online sequential extreme learning machine with kernels for nonstationary time series prediction. Neurocomputing 2014, 145, 90–97. [Google Scholar] [CrossRef]
- Anava, O.; Hazan, E.; Mannor, S.; Shamir, O. Online learning for time series prediction. In Proceedings of the 26th Annual Conference on Learning Theory, Princeton, NJ, USA, 12–14 June 2013; pp. 172–184. [Google Scholar]
- Kuiper, R.M.; Ryan, O. Meta-analysis of lagged regression models: A continuous-time approach. Struct. Equ. Model. A Multidiscip. J. 2020, 27, 396–413. [Google Scholar] [CrossRef] [Green Version]
- Schoenherr, D.; Paulick, J.; Strauss, B.M.; Deisenhofer, A.K.; Schwartz, B.; Rubel, J.A.; Altmann, U. Identification of movement synchrony: Validation of windowed cross-lagged correlation and-regression with peak-picking algorithm. PLoS ONE 2019, 14, e0211494. [Google Scholar] [CrossRef]
- Keele, L.; Kelly, N.J. Dynamic models for dynamic theories: The ins and outs of lagged dependent variables. Political Anal. 2006, 14, 186–205. [Google Scholar] [CrossRef]
- Adhvaryu, P.; Panchal, M. A review on diverse ensemble methods for classification. IOSR J. Comput. Eng. 2012, 1, 27–32. [Google Scholar] [CrossRef]
- Okiy, S.; Nwobi-Okoye, C.C.; Igboanugo, A.C. Transfer Function Modelling: A Literature Survey. Res. J. Appl. Sci. Eng. Technol. 2015, 11, 1265–1279. [Google Scholar] [CrossRef]
- Box, G.E.; Jenkins, G.M.; Reinsel, G.C.; Ljung, G.M. Time Series Analysis: Forecasting and Control; John Wiley & Sons: Hoboken, NJ, USA, 2015. [Google Scholar]
- Davidson, R.; MacKinnon, J.G. Estimation and Inference in Econometrics. OUP Catalogue; Oxford University Press: Oxford, UK, 1993. [Google Scholar]
- Murphy, K.M.; Topel, R.H. Estimation and inference in two-step econometric models. J. Bus. Econ. Stat. 2002, 20, 88–97. [Google Scholar] [CrossRef]
- Timmermann, A. Forecast combinations. Handb. Econ. Forecast. 2006, 1, 135–196. [Google Scholar]
- De Vito, S.; Massera, E.; Piga, M.; Martinotto, L.; Francia, G. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Air+quality (accessed on 28 January 2021).
- De Vito, S.; Massera, E.; Piga, M.; Martinotto, L.; Francia, G. On field calibration of an electronic nose for benzene estimation in an urban pollution monitoring scenario. Sens. Actuators B Chem. 2008, 129, 750–757. [Google Scholar] [CrossRef]
Target Time | Iteration | ARIMA | ARIMAX | ANN | ANNX | RNN | VAR | EDT-r | EDT-w | |
---|---|---|---|---|---|---|---|---|---|---|
1st | time | 0.01 | 5.69 | 0.3 | 3.02 | 2.45 | 0.01 | 0.36 | 1.98 | |
RMSE | 0.5245 | 0.3607 | 0.4705 | 0.3595 | 0.4810 | 0.1356 | 0.7619 | 0.8498 | ||
2nd | time | 0.53 | 4.76 | 0.92 | 18.49 | 4.02 | 0.23 | 0.23 | 0.81 | |
RMSE | 0.5824 | 0.4002 | 0.5217 | 0.4361 | 0.5558 | 0.1815 | 0.8043 | 0.4107 | ||
3rd | time | 0.07 | 0.78 | 0.27 | 2.58 | 0.61 | 0.06 | 0.78 | 0.53 | |
RMSE | 0.5771 | 0.5823 | 0.5660 | 0.5649 | 0.7306 | 0.2495 | 1.4617 | 0.5189 | ||
1st | time | 0.01 | 5.69 | 0.3 | 3.02 | 2.45 | 0.01 | 0.26 | 0.72 | |
RMSE | 0.6163 | 0.4341 | 0.5495 | 0.4451 | 0.4668 | 0.2878 | 0.5642 | 0.3754 | ||
2nd | time | 0.53 | 4.76 | 0.92 | 18.49 | 4.02 | 0.23 | 0.28 | 0.86 | |
RMSE | 0.6348 | 0.4627 | 0.6516 | 0.4902 | 0.5436 | 0.3968 | 0.8547 | 0.1924 | ||
3rd | time | 0.07 | 0.78 | 0.27 | 2.58 | 0.61 | 0.06 | 0.47 | 0.36 | |
RMSE | 0.6241 | 0.7274 | 0.6306 | 0.6436 | 0.7185 | 0.4598 | 0.9037 | 0.2411 | ||
1st | time | 0.01 | 5.69 | 0.3 | 3.02 | 2.45 | 0.01 | 0.22 | 1.13 | |
RMSE | 0.6604 | 0.4191 | 0.6338 | 0.4439 | 0.4637 | 0.3678 | 0.6202 | 0.3311 | ||
2nd | time | 0.53 | 4.76 | 0.92 | 18.49 | 4.02 | 0.23 | 0.32 | 0.75 | |
RMSE | 0.6638 | 0.4747 | 0.6793 | 0.5354 | 0.5469 | 0.4438 | 0.7295 | 0.2470 | ||
3rd | time | 0.07 | 0.78 | 0.27 | 2.58 | 0.61 | 0.06 | 0.26 | 0.38 | |
RMSE | 0.7100 | 0.7961 | 0.7249 | 0.7321 | 0.7208 | 0.6028 | 0.9821 | 0.2087 | ||
1st | time | 0.01 | 5.69 | 0.3 | 3.02 | 2.45 | 0.01 | 0.17 | 1.61 | |
RMSE | 0.6781 | 0.4526 | 0.7137 | 0.5224 | 0.4658 | 0.4245 | 0.6170 | 0.2047 | ||
2nd | time | 0.53 | 4.76 | 0.92 | 18.49 | 4.02 | 0.23 | 1.19 | 2.32 | |
RMSE | 0.6560 | 0.5039 | 0.7386 | 0.5746 | 0.5519 | 0.5420 | 0.6807 | 0.1731 | ||
3rd | time | 0.07 | 0.78 | 0.27 | 2.58 | 0.61 | 0.06 | 0.28 | 0.2 | |
RMSE | 0.7565 | 0.7635 | 0.7958 | 0.7851 | 0.7163 | 0.6555 | 0.9284 | 0.4405 |
Target Time | Iteration | ARIMA | ARIMAX | ANN | ANNX | RNN | VAR | EDT-r | EDT-w | |
---|---|---|---|---|---|---|---|---|---|---|
1st | time | 0.25 | 5.12 | 0.81 | 3.93 | 2.83 | 0.15 | 1.36 | 2.44 | |
RMSE | 0.0049 | 0.0066 | 0.0052 | 0.0065 | 0.2060 | 0.0055 | 0.0589 | 0.0537 | ||
2nd | time | 1.45 | 5.17 | 3.22 | 10.67 | 3.26 | 0.8 | 0.39 | 1.89 | |
RMSE | 0.0415 | 0.0424 | 0.0930 | 0.0502 | 0.5686 | 0.0438 | 0.1612 | 0.0863 | ||
3rd | time | 0.11 | 0.78 | 0.22 | 1.28 | 1.25 | 0.09 | 0.49 | 0.56 | |
RMSE | 0.0322 | 0.0351 | 0.0375 | 0.0433 | 0.3966 | 0.0328 | 0.1368 | 0.0982 | ||
1st | time | 0.25 | 5.12 | 0.81 | 3.93 | 2.83 | 0.15 | 0.65 | 2.31 | |
RMSE | 0.0079 | 0.0092 | 0.0091 | 0.0100 | 0.1646 | 0.0078 | 0.0118 | 0.0040 | ||
2nd | time | 1.45 | 5.17 | 3.22 | 10.67 | 3.26 | 0.8 | 0.28 | 1.48 | |
RMSE | 0.0737 | 0.0827 | 0.1799 | 0.1046 | 0.5683 | 0.0869 | 0.1982 | 0.0943 | ||
3rd | time | 0.11 | 0.78 | 0.22 | 1.28 | 1.25 | 0.09 | 0.37 | 0.6 | |
RMSE | 0.0398 | 0.0530 | 0.0502 | 0.0633 | 0.3935 | 0.0513 | 0.0855 | 0.0501 | ||
1st | time | 0.25 | 5.12 | 0.81 | 3.93 | 2.83 | 0.15 | 0.41 | 2.03 | |
RMSE | 0.0121 | 0.0129 | 0.0147 | 0.0141 | 0.1626 | 0.0108 | 0.0263 | 0.0026 | ||
2nd | time | 1.45 | 5.17 | 3.22 | 10.67 | 3.26 | 0.8 | 0.38 | 3.19 | |
RMSE | 0.1298 | 0.1499 | 0.2638 | 0.1921 | 0.5774 | 0.1630 | 0.4244 | 0.0701 | ||
3rd | time | 0.11 | 0.78 | 0.22 | 1.28 | 1.25 | 0.09 | 0.5 | 0.39 | |
RMSE | 0.0538 | 0.0785 | 0.0720 | 0.0880 | 0.3966 | 0.0796 | 0.1695 | 0.0304 | ||
1st | time | 0.25 | 5.12 | 0.81 | 3.93 | 2.83 | 0.15 | 1.16 | 2.69 | |
RMSE | 0.0193 | 0.0163 | 0.0238 | 0.0186 | 0.1611 | 0.0154 | 0.0330 | 0.0030 | ||
2nd | time | 1.45 | 5.17 | 3.22 | 10.67 | 3.26 | 0.8 | 0.5 | 1.38 | |
RMSE | 0.2367 | 0.2190 | 0.3531 | 0.2828 | 0.5928 | 0.2470 | 0.3131 | 0.0794 | ||
3rd | time | 0.11 | 0.78 | 0.22 | 1.28 | 1.25 | 0.09 | 0.43 | 0.44 | |
RMSE | 0.0747 | 0.1216 | 0.1038 | 0.1303 | 0.4093 | 0.1249 | 0.2065 | 0.0406 |
Target Time | Iteration | ARIMA | ARIMAX | ANN | ANNX | RNN | VAR | EDT-r | EDT-w | |
---|---|---|---|---|---|---|---|---|---|---|
5000 to 6000 | time | 0.06 | 0.71 | 0.22 | 2.79 | 0.51 | 0.04 | 0.47 | 0.53 | |
RMSE | 0.4282 | 0.4259 | 0.4312 | 0.4134 | 1.1139 | 0.4561 | 0.6609 | 0.4821 | ||
6000 to 7000 | time | 0.03 | 0.7 | 0.25 | 1.86 | 0.77 | 0.02 | 0.29 | 0.41 | |
RMSE | 0.3346 | 0.3394 | 0.3537 | 0.3352 | 0.9868 | 0.3579 | 0.4570 | 0.2243 | ||
7000 to 8000 | time | 0.05 | 0.47 | 0.31 | 1.34 | 0.56 | 0.03 | 0.37 | 0.6 | |
RMSE | 0.3039 | 0.3067 | 0.3194 | 0.3082 | 0.9522 | 0.3215 | 0.5261 | 0.3211 | ||
8000 to 9000 | time | 0.07 | 0.73 | 0.8 | 7.7 | 0.46 | 0.04 | 0.3 | 0.56 | |
RMSE | 0.3025 | 0.3066 | 0.2681 | 0.2545 | 0.8133 | 0.2956 | 0.7946 | 0.1912 | ||
5000 to 6000 | time | 0.06 | 0.71 | 0.22 | 2.79 | 0.51 | 0.04 | 0.47 | 0.41 | |
RMSE | 0.8377 | 0.8562 | 0.8559 | 0.7954 | 1.1069 | 0.4967 | 1.7584 | 0.3451 | ||
6000 to 7000 | time | 0.03 | 0.7 | 0.25 | 1.86 | 0.77 | 0.02 | 0.39 | 0.48 | |
RMSE | 0.6535 | 0.6603 | 0.7072 | 0.6495 | 0.9827 | 0.6822 | 1.6410 | 0.2156 | ||
7000 to 8000 | time | 0.05 | 0.47 | 0.31 | 1.34 | 0.56 | 0.03 | 0.4 | 0.53 | |
RMSE | 0.5941 | 0.6021 | 0.6037 | 0.5978 | 0.9446 | 0.6252 | 1.1272 | 0.1168 | ||
8000 to 9000 | time | 0.07 | 0.73 | 0.8 | 7.7 | 0.46 | 0.04 | 0.34 | 0.59 | |
RMSE | 0.5580 | 0.5841 | 0.4508 | 0.4465 | 0.7882 | 0.5384 | 1.3260 | 0.1284 | ||
5000 to 6000 | time | 0.06 | 0.71 | 0.22 | 2.79 | 0.51 | 0.04 | 0.34 | 0.53 | |
RMSE | 1.0202 | 1.0590 | 1.0906 | 1.0282 | 1.1181 | 1.0086 | 1.1211 | 0.5270 | ||
6000 to 7000 | time | 0.03 | 0.7 | 0.25 | 1.86 | 0.77 | 0.02 | 0.37 | 0.52 | |
RMSE | 0.8521 | 0.8711 | 0.9710 | 0.8972 | 0.9829 | 0.8385 | 0.8362 | 0.2502 | ||
7000 to 8000 | time | 0.05 | 0.47 | 0.31 | 1.34 | 0.56 | 0.03 | 0.54 | 0.48 | |
RMSE | 0.7823 | 0.7959 | 0.7716 | 0.7597 | 0.9463 | 0.7899 | 0.7107 | 0.2854 | ||
8000 to 9000 | time | 0.07 | 0.73 | 0.8 | 7.7 | 0.46 | 0.04 | 0.45 | 0.57 | |
RMSE | 0.6162 | 0.6493 | 0.5382 | 0.5099 | 0.7919 | 0.6114 | 0.6127 | 0.1762 | ||
5000 to 6000 | time | 0.06 | 0.71 | 0.22 | 2.79 | 0.51 | 0.04 | 0.31 | 0.41 | |
RMSE | 1.0553 | 1.0776 | 1.1929 | 1.0921 | 1.1187 | 1.2360 | 0.9430 | 0.4948 | ||
6000 to 7000 | time | 0.03 | 0.7 | 0.25 | 1.86 | 0.77 | 0.02 | 0.36 | 0.68 | |
RMSE | 0.9182 | 0.9518 | 1.1717 | 1.0377 | 0.9808 | 0.9636 | 0.8282 | 0.2899 | ||
7000 to 8000 | time | 0.05 | 0.47 | 0.31 | 1.34 | 0.56 | 0.03 | 0.59 | 0.6 | |
RMSE | 0.8672 | 0.8716 | 0.8701 | 0.8208 | 0.9563 | 0.9639 | 0.7210 | 0.4015 | ||
8000 to 9000 | time | 0.07 | 0.73 | 0.8 | 7.7 | 0.46 | 0.04 | 0.5 | 0.54 | |
RMSE | 0.6122 | 0.6393 | 0.5721 | 0.5411 | 0.7987 | 0.5785 | 0.6455 | 0.2702 |
Log Data | Description |
---|---|
Used memory | memory occupancy for a target application |
System CPU | CPU occupancy for a target application |
User CPU | CPU occupancy for all application |
Count | The number of simultaneous access |
Num http | The number of thread for a target application about http requests |
Num javad | The number of thread for a target application about javad |
Num jdbc | The number of thread for a target application about jdbc |
Response | Response time of a target application to a request of users |
Target Time | Iteration | ARIMA | ARIMAX | ANN | ANNX | RNN | VAR | EDT-r | EDT-w | |
---|---|---|---|---|---|---|---|---|---|---|
1st non-stationary | time | 0.16 | 4.23 | 0.25 | 2.39 | 1.75 | 0.06 | 1.94 | 1.14 | |
RMSE | 0.1045 | 0.0412 | 0.0232 | 0.0423 | 0.3434 | 0.0107 | 0.1003 | 0.0909 | ||
2nd non-stationary | time | 0.19 | 6.43 | 0.13 | 14.96 | 2.36 | 0.11 | 0.72 | 0.72 | |
RMSE | 0.0263 | 0.0200 | 0.0100 | 0.0310 | 0.3071 | 0.0040 | 0.0511 | 0.0348 | ||
3rd non-stationary | time | 0.1 | 1.36 | 0.18 | 1.6 | 0.51 | 0.01 | 0.34 | 0.44 | |
RMSE | 0.1038 | 0.3772 | 0.1168 | 0.1508 | 1.3540 | 0.0507 | 0.4286 | 0.3403 | ||
4th non-stationary | time | 0.14 | 1.34 | 0.31 | 1.3 | 0.45 | 0.02 | 0.39 | 0.23 | |
RMSE | 0.1457 | 0.1297 | 0.0312 | 0.0628 | 0.4285 | 0.0153 | 0.1344 | 0.1257 | ||
1st stationary | time | 0.14 | 1.27 | 0.24 | 1.67 | 0.78 | 0.01 | 0.27 | 0.37 | |
RMSE | 0.0014 | 0.0014 | 0.0002 | 0.0003 | 0.1555 | 0.0003 | 0.0075 | 0.0075 | ||
2nd stationary | time | 0.14 | 1.28 | 0.13 | 1.43 | 1.22 | 0.08 | 0.38 | 0.34 | |
RMSE | 0.0018 | 0.0016 | 0.0003 | 0.0003 | 0.1641 | 0.0003 | 0.0081 | 0.0082 | ||
3rd stationary | time | 0.09 | 1.28 | 0.16 | 1.59 | 2.72 | 0.05 | 0.37 | 1.69 | |
RMSE | 0.0014 | 0.0013 | 0.0003 | 0.0004 | 0.1651 | 0.0003 | 0.0076 | 0.0076 | ||
4th stationary | time | 0.15 | 1.04 | 0.2 | 1.64 | 0.35 | 0.07 | 0.18 | 0.48 | |
RMSE | 0.0013 | 0.0011 | 0.0004 | 0.0005 | 0.1710 | 0.0005 | 0.0077 | 0.0077 | ||
1st non-stationary | time | 0.16 | 4.23 | 0.25 | 2.39 | 1.75 | 0.06 | 0.65 | 2.24 | |
RMSE | 0.1693 | 0.1313 | 0.1570 | 0.2012 | 0.2859 | 0.1280 | 0.1662 | 0.1558 | ||
2nd non-stationary | time | 0.19 | 6.43 | 0.13 | 14.96 | 2.36 | 0.11 | 0.17 | 0.56 | |
RMSE | 0.0716 | 0.0644 | 0.0839 | 0.0898 | 0.2557 | 0.0404 | 0.1363 | 0.1286 | ||
3rd non-stationary | time | 0.1 | 1.36 | 0.18 | 1.6 | 0.51 | 0.01 | 0.2 | 0.56 | |
RMSE | 0.4501 | 0.6450 | 0.7157 | 0.7266 | 1.2954 | 0.6688 | 1.1211 | 1.0639 | ||
4th non-stationary | time | 0.14 | 1.34 | 0.31 | 1.3 | 0.45 | 0.02 | 0.36 | 0.27 | |
RMSE | 0.2325 | 0.2791 | 0.2083 | 0.2899 | 0.3510 | 0.1928 | 0.2118 | 0.1886 | ||
1st stationary | time | 0.14 | 1.27 | 0.24 | 1.67 | 0.78 | 0.01 | 0.14 | 0.42 | |
RMSE | 0.0060 | 0.0064 | 0.0022 | 0.0026 | 0.1135 | 0.0024 | 0.0130 | 0.0112 | ||
2nd stationary | time | 0.14 | 1.28 | 0.13 | 1.43 | 1.22 | 0.08 | 0.18 | 0.43 | |
RMSE | 0.0071 | 0.0061 | 0.0025 | 0.0029 | 0.1167 | 0.0024 | 0.0128 | 0.0122 | ||
3rd stationary | time | 0.09 | 1.28 | 0.16 | 1.59 | 2.72 | 0.05 | 0.53 | 2.7 | |
RMSE | 0.0053 | 0.0050 | 0.0026 | 0.0030 | 0.1170 | 0.0025 | 0.0230 | 0.0237 | ||
4th stationary | time | 0.15 | 1.04 | 0.2 | 1.64 | 0.35 | 0.07 | 0.19 | 0.43 | |
RMSE | 0.0049 | 0.0046 | 0.0028 | 0.0034 | 0.1197 | 0.0033 | 0.0174 | 0.0171 | ||
1st non-stationary | time | 0.16 | 4.23 | 0.25 | 2.39 | 1.75 | 0.06 | 1.35 | 0.51 | |
RMSE | 0.2617 | 0.3144 | 0.3208 | 0.3204 | 0.2897 | 0.8960 | 0.3072 | 0.1404 | ||
2nd non-stationary | time | 0.19 | 6.43 | 0.13 | 14.96 | 2.36 | 0.11 | 0.27 | 0.6 | |
RMSE | 0.1470 | 0.1287 | 0.1394 | 0.1311 | 0.2557 | 0.2262 | 0.1141 | 0.0519 | ||
3rd non-stationary | time | 0.1 | 1.36 | 0.18 | 1.6 | 0.51 | 0.01 | 0.16 | 0.27 | |
RMSE | 1.2191 | 1.0748 | 1.5719 | 1.8800 | 1.2983 | 7.4417 | 1.2883 | 1.1557 | ||
4th non-stationary | time | 0.14 | 1.34 | 0.31 | 1.3 | 0.45 | 0.02 | 0.28 | 0.25 | |
RMSE | 0.3511 | 0.5330 | 0.4426 | 0.3829 | 0.3500 | 1.6100 | 0.2739 | 0.2432 | ||
1st stationary | time | 0.14 | 1.27 | 0.24 | 1.67 | 0.78 | 0.01 | 0.25 | 0.35 | |
RMSE | 0.0144 | 0.0150 | 0.0089 | 0.0099 | 0.1130 | 0.0092 | 0.0155 | 0.0080 | ||
2nd stationary | time | 0.14 | 1.28 | 0.13 | 1.43 | 1.22 | 0.08 | 0.3 | 0.39 | |
RMSE | 0.0160 | 0.0136 | 0.0095 | 0.0111 | 0.1154 | 0.0092 | 0.0182 | 0.0087 | ||
3rd stationary | time | 0.09 | 1.28 | 0.16 | 1.59 | 2.72 | 0.05 | 2.16 | 2.58 | |
RMSE | 0.0128 | 0.0118 | 0.0098 | 0.0109 | 0.1186 | 0.0091 | 0.0185 | 0.0090 | ||
4th stationary | time | 0.15 | 1.04 | 0.2 | 1.64 | 0.35 | 0.07 | 0.28 | 0.23 | |
RMSE | 0.0125 | 0.0122 | 0.0104 | 0.0117 | 0.1175 | 0.0129 | 0.0183 | 0.0101 | ||
1st non-stationary | time | 0.16 | 4.23 | 0.25 | 2.39 | 1.75 | 0.06 | 0.67 | 0.49 | |
RMSE | 0.3718 | 0.5571 | 0.5494 | 0.5292 | 0.2907 | 14.6659 | 0.2539 | 0.2141 | ||
2nd non-stationary | time | 0.19 | 6.43 | 0.13 | 14.96 | 2.36 | 0.11 | 0.39 | 0.68 | |
RMSE | 0.2263 | 0.1813 | 0.2871 | 0.1649 | 0.2607 | 2.4823 | 0.1190 | 0.1212 | ||
3rd non-stationary | time | 0.1 | 1.36 | 0.18 | 1.6 | 0.51 | 0.01 | 0.18 | 0.36 | |
RMSE | 2.7556 | 1.8032 | 4.0593 | 6.1531 | 1.2995 | 4286.3025 | 1.5128 | 1.8513 | ||
4th non-stationary | time | 0.14 | 1.34 | 0.31 | 1.3 | 0.45 | 0.02 | 0.21 | 0.39 | |
RMSE | 0.4862 | 0.8009 | 0.5762 | 0.5887 | 0.3554 | 46.5682 | 0.3411 | 0.2814 | ||
1st stationary | time | 0.14 | 1.27 | 0.24 | 1.67 | 0.78 | 0.01 | 0.24 | 0.31 | |
RMSE | 0.0215 | 0.0203 | 0.0169 | 0.0191 | 0.1131 | 0.0154 | 0.0154 | 0.0144 | ||
2nd stationary | time | 0.14 | 1.28 | 0.13 | 1.43 | 1.22 | 0.08 | 0.37 | 0.36 | |
RMSE | 0.0212 | 0.0190 | 0.0167 | 0.0184 | 0.1180 | 0.0158 | 0.0156 | 0.0143 | ||
3rd stationary | time | 0.09 | 1.28 | 0.16 | 1.59 | 2.72 | 0.05 | 1.35 | 1.45 | |
RMSE | 0.0182 | 0.0179 | 0.0177 | 0.0185 | 0.1167 | 0.0158 | 0.0172 | 0.0152 | ||
4th stationary | time | 0.15 | 1.04 | 0.2 | 1.64 | 0.35 | 0.07 | 0.2 | 0.27 | |
RMSE | 0.0185 | 0.0190 | 0.0185 | 0.0206 | 0.1162 | 0.8072 | 0.0166 | 0.0159 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Kong, T.; Choi, D.; Lee, G.; Lee, K. Air Pollution Prediction Using an Ensemble of Dynamic Transfer Models for Multivariate Time Series. Sustainability 2021, 13, 1367. https://doi.org/10.3390/su13031367
Kong T, Choi D, Lee G, Lee K. Air Pollution Prediction Using an Ensemble of Dynamic Transfer Models for Multivariate Time Series. Sustainability. 2021; 13(3):1367. https://doi.org/10.3390/su13031367
Chicago/Turabian StyleKong, Taewoon, Dongguen Choi, Geonseok Lee, and Kichun Lee. 2021. "Air Pollution Prediction Using an Ensemble of Dynamic Transfer Models for Multivariate Time Series" Sustainability 13, no. 3: 1367. https://doi.org/10.3390/su13031367