Application of Recurrent Neural Networks in time series of meteorological stations for data imputation: an approach on meteorological micro-stations in the Western region of Pará
Abstract
The multivariate data of the temporary series are present in a large number of applications, and in many cases the lost of information at historical series is a recurrint problem. Based on this problem, this work presents models that use recurrent neural networks for data imputation on multivariate series of meteorological stations, in this case listed at the western region of Pará. The dataset used presented consists in historical series with meteorological variables that have large data gaps. The results obtained in this work present two models of recurrent neural networks that demonstrate how it is possible to perform the imputation of data in multivariate time series for each meteorological variable individually. An approach is also developed comparing models of LSTM and GRU networks to demonstrate the efficiency of recurrent networks as an alternative to the imputation of multivariate data in time series of weather stations.
References
Che, Z., Purushotham, S., Cho, K., Sontag, D., and Liu, Y. (2018). Recurrent neural networks for multivariate time series with missing values. Scientific reports, 8(1):6085.
Chung, J., Gulcehre, C., Cho, K., and Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555.
Costa, R. L., Silva, F., Sarmanho, G. F., and Lucio, P. S. (2012). Imputação multivariada de dados diários de precipitação e análise de índices de extremos climáticos. Revista Brasileira de Geografia Física, 3:661–675.
Dey, R. and Salemt, F. M. (2017). Gate-variants of gated recurrent unit (gru) neural networks. In 2017 IEEE 60th international midwest symposium on circuits and systems (MWSCAS), pages 1597–1600. IEEE.
Enders, C. K. (2010). Applied missing data analysis. Guilford press.
Ferreira, J., Tapajós, R., and Conde, G. (2016). Redes neurais artificiais para o preenchimento de falhas em séries temporais meteorológicas.
Gensler, A., Henze, J., Sick, B., and Raabe, N. (2016). Deep learning for solar power forecasting— an approach using autoencoder and lstm neural networks. In 2016 IEEE international conference on systems, man, and cybernetics (SMC), pages 002858–002865. IEEE.
Graves, A., Liwicki, M., Fernández, S., Bertolami, R., Bunke, H., and Schmidhuber, J. (2008). A novel connectionist system for unconstrained handwriting recognition. IEEE transactions on pattern analysis and machine intelligence, 31(5):855–868.
Kim, H.-G., Jang, G.-J., Choi, H.-J., Kim, M., Kim, Y.-W., and Choi, J. (2017). Recurrent neural networks with missing information imputation for medical examination data prediction. In 2017 IEEE International Conference on Big Data and Smart Computing (BigComp), pages 317–323. IEEE.
Little, R. J. and Rubin, D. B. (2019). Statistical analysis with missing data, volume 793. John Wiley & Sons.
Schafer, J. L. (1997). Analysis of incomplete multivariate data. Chapman and Hall/CRC.
Souza, R. (1981). Metodologias para a análise e previsão de séries temporais univariadas e multivariadas. Brazilian Review of Econometrics, 1(2):78–105.
Wilks, S. S. (1932). Moments and distributions of estimates of population parameters from fragmentary samples. The Annals of Mathematical Statistics, 3(3):163–195.
