Gap filling in air temperature time series: a comparison between machine learning models
DOI:
https://doi.org/10.55761/abclima.v35i20.17649Keywords:
Imputing missing values. Machine Learning. Decision Trees. Support Vector Machines. SVR. CART. Elastic Net. LASSO. KNN. Linear Regression.Abstract
This study conducted a comparative analysis of different machine learning (ML) algorithms for filling gaps in air temperature data from four different locations in Brazilian states. Six algorithms were evaluated: linear regression, LASSO regression, elastic net, k-nearest neighbors, decision trees (CART), and support vector regression (SVR). The results, covering all sites, indicate that the Support Vector Regression (SVR) model was the most promising, with exceptionally low RMSE ranging from 0.1712 °C to 0.2062 °C. This suggests that SVR may be the best choice for predicting air temperature. While Decision Trees showed robust results with RMSE ranging from 0.2198 °C to 0.3746 °C. The Elastic Net (EN) and LASSO models performed poorly, with RMSE ranging from 1.6935 °C to 2.8555 °C. The K-Nearest Neighbors (KNN) model produced intermediate results, with RMSE ranging from 0.5579 °C to 0.7567 °C. Linear Regression also showed variable results, with RMSE ranging from 0.7474 °C to 1.4010 °C.
Downloads
References
AWAD, Mariette; KHANNA, Rahul. Efficient learning machines: theories, concepts, and applications for engineers and system designers. Springer nature, 2015.
BONFANTE, Andreia Gentil et al. Uma abordagem computacional para preenchimento de falhas em dados micro meteorológicos. Revista Brasileira de Ciências Ambientais (RBCIAMB), n. 27, p. 61-70, 2013.
BREIMAN, Leo et al. Classification and regression trees. CRC press, 1984.
CONNELLY, Lynne. Logistic regression. Medsurg Nursing, v. 29, n. 5, p. 353-354, 2020.
COULIBALY, P.; EVORA, N. D. Comparison of neural network methods for infilling missing daily weather records. Journal of hydrology, v. 341, n. 1-2, p. 27-41, 2007.
CHATTERJEE, Soumyadeep et al. Sparse group lasso for regression on land climate variables. In: 2011 IEEE 11th International Conference on Data Mining Workshops. IEEE, 2011. p. 1-8.
FIX, Evelyn; HODGES, Joseph Lawson. Discriminatory analysis. Nonparametric discrimination: Consistency properties. International Statistical Review/Revue Internationale de Statistique, v. 57, n. 3, p. 238-247, 1989.
FRIEDMAN, Jerome; HASTIE, Trevor; TIBSHIRANI, Rob. Regularization paths for generalized linear models via coordinate descent. Journal of statistical software, v. 33, n. 1, p. 1, 2010.
GARRETA, Raul; MONCECCHI, Guillermo. Learning scikit-learn: machine learning in python. Packt Publishing Ltd, 2013.
KAJEWSKA-SZKUDLAREK, Joanna; STAŃCZYK, Justyna. Filling missing meteorological data with Computational Intelligence methods. In: ITM web of conferences. EDP Sciences, 2018. p. 00015.
KATİPOĞLU, Okan Mert; REŞAT, A. C. A. R. Estimation of missing temperature data by Artificial Neural Network (ANN). Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi, v. 12, n. 2, p. 431-438, 2021.
LATIF, Sarmad Dashti et al. Assessing rainfall prediction models: Exploring the advantages of machine learning and remote sensing approaches. Alexandria Engineering Journal, v. 82, p. 16-25, 2023.
MATSUMOTO, Makoto; NISHIMURA, Takuji. Mersenne twister: a 623-dimensionally equidistributed uniform pseudo-random number generator. ACM Transactions on Modeling and Computer Simulation (TOMACS), v. 8, n. 1, p. 3-30, 1998.
MEGETO, Guilherme AS et al. Decision tree for classification of soybean rust occurence in commercial crops based on weather variables. Engenharia Agrícola, v. 34, p. 590-599, 2014.
MOHAMMADI, Kasra et al. Extreme learning machine based prediction of daily dew point temperature. Computers and Electronics in Agriculture, v. 117, p. 214-225, 2015.
MORI, Hiroyuki; TAKAHASHI, Akira. A data mining method for selecting input variables for forecasting model of global solar radiation. In: PES T&D 2012. IEEE, 2012. p. 1-6.
MORRIS, Clint; YANG, Jidong J. Effectiveness of resampling methods in coping with imbalanced crash data: Crash type analysis and predictive modeling. Accident Analysis & Prevention, v. 159, p. 106240, 2021.
PATRICK, Edward A.; FISCHER III, Frederic P. A generalized k-nearest neighbor rule. Information and control, v. 16, n. 2, p. 128-152, 1970.
PEDRO, Hugo TC; COIMBRA, Carlos FM. Nearest-neighbor methodology for prediction of intra-hour global horizontal and direct normal irradiances. Renewable Energy, v. 80, p. 770-782, 2015.
RAOUHI, El Mehdi; LACHGAR, Mohamed; KARTIT, Ali. Comparative Study of Regression and Regularization Methods: Application to Weather and Climate Data. In: WITS 2020: Proceedings of the 6th International Conference on Wireless Technologies, Embedded, and Intelligent Systems. Springer Singapore, 2022. p. 233-240.
SMOLA, Alex J.; SCHÖLKOPF, Bernhard. A tutorial on support vector regression. Statistics and computing, v. 14, p. 199-222, 2004.
TCHAKONTE, Siméon et al. Using machine learning models to assess the population dynamic of the freshwater invasive snail Physa acuta Draparnaud, 1805 (Gastropoda: Physidae) in a tropical urban polluted streams-system. Limnologica, v. 99, p. 126049, 2023.
THOBER, Stephan et al. Multi-model ensemble projections of European river floods and high flows at 1.5, 2, and 3 degrees global warming. Environmental Research Letters, v. 13, n. 1, p. 014003, 2018.
TOSUNOĞLU, Fatih et al. Monthly streamflow forecasting using machine learning. Erzincan University Journal of Science and Technology, v. 13, n. 3, p. 1242-1251, 2020.
WEN, Jiabao et al. Big data driven marine environment information forecasting: a time series prediction network. IEEE Transactions on Fuzzy Systems, v. 29, n. 1, p. 4-18, 2020.
XU, Yongjun et al. Artificial intelligence: A powerful paradigm for scientific research. The Innovation, v. 2, n. 4, 2021.
Downloads
Published
How to Cite
Issue
Section
License
A aprovação dos artigos implica a aceitação imediata e sem ônus de que a Revista Brasileira de Climatologia terá exclusividade na primeira publicação do artigo. Os autores continuarão, não obstante, a deter os direitos autorais. Os autores autorizam também que seus artigos sejam disponibilizados em todos os indexadores aos quais a revista está vinculada.
Os autores mantém seus direitos de publicação sem restrições
A Comissão Editorial não se responsabiliza pelos conceitos ou afirmações expressos nos trabalhos publicados, que são de inteira responsabilidade dos autores.
A Revista Brasileira de Climatologia oferece acesso livre imediato ao seu conteúdo, seguindo o entendimento de que disponibilizar gratuitamente o conhecimento científico ao público proporciona maior democratização do conhecimento e tende a produzir maior impacto dos artigos publicados. Os artigos publicados na revista são disponibilizados segundo a Licença Creative Commons CC-BY-NC 4.0 (https://creativecommons.org/licenses/by-nc/4.0/). Segundo essa licença é permitido acessar, distribuir e reutilizar os artigos para fins não comerciais desde que citados os autores e a fonte. Ao submeter artigos à Revista Brasileira de Climatologia, os autores concordam em tornar seus textos legalmente disponíveis segundo essa licença