A Comparative Evaluation of Various Imputation Methods for LSTM-Based Climatological Time Series Forecasting
Keywords:
Imputation, Missing Values, LSTM, Forecasting, ClimatologyAbstract
Missing values substantially degrade the reliability of environmental time-series forecasting; however, prior studies largely evaluate imputation methods in isolation without systematically linking missingness mechanisms to deep learning forecasting performance. To address this gap, this study proposes a mechanism-aware comparative framework that evaluates deletion and six imputation methods (Mean, Median, Mode, LOCF, KNN, and MICE) across three environmental time-series datasets with naturally occurring missing values, using LSTM as the forecasting model. The novelty lies in jointly analyzing statistical error (MAPE, RMSE), goodness-of-fit (R²), and statistical significance to identify structurally aligned imputation strategies under different missingness patterns. Experimental results show that deletion as baseline consistently produces the worst performance (MAPE: 5.91429; 7.35000; 2.84881), whereas imputation reduces proportional error by more than 70% on average (p < 0.05). LOCF performs best under temporal dependency (MAPE 0.73959; R² 0.92757), KNN achieves the most balanced performance under MCAR-like behavior (R² 0.94086), and Mean imputation yields the lowest error in MAR-structured data (MAPE 0.41560; R² 0.97077). These findings demonstrate that imputation effectiveness depends on alignment with missingness structure rather than methodological complexity, providing evidence-based guidance for robust environmental.
References
[1] M. Patacca, M. Lindner, M. E. Lucas-Borja, T. Cordonnier, G. Fidej, B. Gardiner, Y. Hauf, G. Jasinevičius, S. Labonne, E. Linkevičius, M. Mahnken, S. Milanovic, G. Nabuurs, T. A. Nagel, L. Nikinmaa, M. Panyatov, R. Bercak, R. Seidl, M. Z. Ostrogović Sever, J. Socha, D. Thom, D. Vuletic, S. Zudin, and M. Schelhaas, “Significant increase in natural disturbance impacts on European forests since 1950,” Global Change Biology, vol. 29, no. 5, pp. 1359–1376, Mar. 2023, doi: 10.1111/gcb.16531.
[2] U. A. Usmani, I. Abdul Aziz, J. Jaafar, and J. Watada, “Deep Learning for Anomaly Detection in Time-Series Data: An Analysis of Techniques, Review of Applications, and Guidelines for Future Research,” IEEE Access, vol. 12, pp. 174564–174590, 2024, doi: 10.1109/ACCESS.2024.3495819.
[3] K. Zhang, Q. Yang, C. Li, X. Sun, and J. Chen, “Missing Data Recovery Methods on Multivariate Time Series in IoT: A Comprehensive Survey,” IEEE Commun. Surv. Tutorials, vol. 28, pp. 149–180, 2026, doi: 10.1109/COMST.2025.3585962.
[4] Z. Li, J. F. L. Chavez, S. Pant, and D. Li, “Enhancing Ecological Forecasting with LSTM Models: The Impact of Partition-Based Data Shuffling on Predictive Accuracy,” in SoutheastCon 2025, Concord, NC, USA: IEEE, Mar. 2025, pp. 314–319. doi: 10.1109/SoutheastCon56624.2025.10971684.
[5] D. Popovich, “How To Treat Missing Data In Survey Research,” Journal of Marketing Theory and Practice, vol. 33, no. 1, pp. 43–59, Jan. 2025, doi: 10.1080/10696679.2024.2376052.
[6] A. D. Woods, D. Gerasimova, B. V. Dusen, J. Nissen, S. Bainter, A. Uzdavines, P. E. Davis-Kean, M. Halvorson, K. M. King, J. A. R. Logan, M. Xu, M. R. Vasilev, J. M. Clay, D. Moreau, K. Joyal-Desmarais, R. A. Cruz, D. M. Y. Brown, K. Schmidt, and M. M. Elsherif, “Best practices for addressing missing data through multiple imputation,” Infant and Child Development, vol. 33, no. 1, pp. 1–37, Jan. 2024, doi: 10.1002/icd.2407.
[7] H. Wang, Z. Chen, Y. Shen, H. Zheng, D. Yang, D. Zhao, and B. Liang, “Robust Missing Value Imputation With Proximal Optimal Transport for Low-Quality IIoT Data,” IEEE Trans. Neural Netw. Learning Syst., vol. 37, no. 1, pp. 82–94, Jan. 2026, doi: 10.1109/TNNLS.2025.3601130.
[8] D. Mavridis, G. Salanti, T. A. Furukawa, A. Cipriani, A. Chaimani, and I. R. White, “Allowing for uncertainty due to missing and LOCF imputed outcomes in meta‐analysis,” Statistics in Medicine, vol. 38, no. 5, pp. 720–737, Feb. 2019, doi: 10.1002/sim.8009.
[9] L. Bourguignon, L. P. Lukas, J. D. Guest, F. H. Geisler, V. Noonan, A. Curt, S. C. Brüningk, and C. R. Jutzeler, “Studying missingness in spinal cord injury data: challenges and impact of data imputation,” BMC Med Res Methodol, vol. 24, no. 1, p. 5, Jan. 2024, doi: 10.1186/s12874-023-02125-x.
[10] M. Rizki, A. Hermawan, and D. Avianto, “Optimization of Hyperparameter K in K-Nearest Neighbor Using Particle Swarm Optimization,” JUITA, vol. 12, no. 1, p. 71, May 2024, doi: 10.30595/juita.v12i1.20688.
[11] A. Altamimi, A. A. Alarfaj, M. Umer, E. A. Alabdulqader, S. Alsubai, T. Kim, and I. Ashraf, “An automated approach to predict diabetic patients using KNN imputation and effective data mining techniques,” BMC Med Res Methodol, vol. 24, no. 1, p. 221, Sep. 2024, doi: 10.1186/s12874-024-02324-0.
[12] K. P. Junaid, T. Kiran, M. Gupta, K. Kishore, and S. Siwatch, “How much missing data is too much to impute for longitudinal health indicators? A preliminary guideline for the choice of the extent of missing proportion to impute with multiple imputation by chained equations,” Popul Health Metrics, vol. 23, no. 1, p. 2, Feb. 2025, doi: 10.1186/s12963-025-00364-2.
[13] Z. Jinbo, L. Yufu, and M. Haitao, “Handling missing data of using the XGBoost-based multiple imputation by chained equations regression method,” Front. Artif. Intell., vol. 8, p. 1553220, Apr. 2025, doi: 10.3389/frai.2025.1553220.
[14] M. Boukdire, Ç. A. İnan, G. Varra, R. Della Morte, and L. Cozzolino, “Interpolation and Machine Learning Methods for Sub-Hourly Missing Rainfall Data Imputation in a Data-Scarce Environment: One- and Two-Step Approaches,” Hydrology, vol. 12, no. 11, p. 297, Nov. 2025, doi: 10.3390/hydrology12110297.
[15] Alfarelzi, M. A. Hakim, M. R. Ediananta, B. A. Pramudita, F. Corputty, and D. P. Setiawan, “Imputation of Missing Values for Generating Rainfall Data Using a Multi-Method Approach,” in 2025 IEEE 16th Control and System Graduate Research Colloquium (ICSGRC), Shah Alam, Malaysia: IEEE, Aug. 2025, pp. 143–148. doi: 10.1109/ICSGRC65918.2025.11159759.
[16] W. Pangesti, N. Syukri, K. A. Notodiputro, Y. Angraini, and L. N. A. Mualifah, “Performance Evaluation of ARIMA and GRU Models for Forecasting Chili Price in East Jawa,” JUITA, vol. 13, no. 2, pp. 209–218, Aug. 2025, doi: 10.30595/juita.v13i2.26445.
[17] N. Niako, J. D. Melgarejo, G. E. Maestre, and K. P. Vatcheva, “Effects of missing data imputation methods on univariate blood pressure time series data analysis and forecasting with ARIMA and LSTM,” BMC Med Res Methodol, vol. 24, no. 1, p. 320, Dec. 2024, doi: 10.1186/s12874-024-02448-3.
[18] M. L. Hossain, S. M. N. Shams, and S. M. Ullah, “Time-series and deep learning approaches for renewable energy forecasting in Dhaka: a comparative study of ARIMA, SARIMA, and LSTM models,” Discov Sustain, vol. 6, no. 1, p. 775, Aug. 2025, doi: 10.1007/s43621-025-01733-5.
[19] A. Rouabhi, “Century-Scale Climate Evolution in Semiarid High Plateaus, Algeria,” Journal of Applied Meteorology and Climatology, vol. 65, no. 2, pp. 311–321, Feb. 2026, doi: 10.1175/JAMC-D-25-0181.1.
[20] M. A. De Sousa Guilhon Araujo, F. L. Cyrino Oliveira, and R. C. Souza, “Missing data imputation using climate data reanalysis in solar energy generation time series,” Energy Conversion and Management: X, vol. 30, p. 101653, May 2026, doi: 10.1016/j.ecmx.2026.101653.
[21] C. Wongoutong, “Performance Comparison of Various Imputation Methods for Missing Data Mechanisms (MAR, MCAR, and MNAR) in a Nonstationary Time Series,” International Journal of Mathematics and Mathematical Sciences, vol. 2025, no. 1, p. 3031708, Jan. 2025, doi: 10.1155/ijmm/3031708.
[22] F. T. Bahadur, S. R. Shah, and R. R. Nidamanuri, “Geospatial time series forecasting of air quality across Delhi using ARIMA and LSTM models,” Journal of Atmospheric and Solar-Terrestrial Physics, vol. 279, p. 106723, Feb. 2026, doi: 10.1016/j.jastp.2026.106723.
[23] W. Astuti and R. Govindaraju, “Effect of Data Gaps on Harmful Algal Bloom Prediction Models for Inland Lakes,” J. Hydrol. Eng., vol. 31, no. 1, p. 04025045, Feb. 2026, doi: 10.1061/JHYEFF.HEENG-6544.
[24] S. Khalili and H. Chulia, “Data Imputation in Large Datasets: A Comparative Study of PCA and Machine Learning Approaches,” Comput Econ, Jan. 2026, doi: 10.1007/s10614-025-11201-x.
[25] H. Gong and K. W. Chan, “Sensitivity analysis for generalized estimating equation with non‐ignorable missing data,” Scandinavian J Statistics, p. sjos.70060, Feb. 2026, doi: 10.1111/sjos.70060.
[26] Y. Zhang and P. J. Thorburn, “A Dual-Head Attention Model for Time Series Data Imputation,” Computers and Electronics in Agriculture, vol. 189, p. 106377, Oct. 2021, doi: 10.1016/j.compag.2021.106377.
[27] A. Palanivinayagam and R. Damaševičius, “Effective Handling of Missing Values in Datasets for Classification Using Machine Learning Methods,” Information, vol. 14, no. 2, p. 92, Feb. 2023, doi: 10.3390/info14020092.
[28] A. Almeida, S. Brás, S. Sargento, and F. C. Pinto, “Focalize K-NN: an imputation algorithm for time series datasets,” Pattern Anal Applic, vol. 27, no. 2, p. 39, Jun. 2024, doi: 10.1007/s10044-024-01262-3.
[29] M. Muhammad, T. Sutikno, and I. Riadi, “A Comparative Study of K-Means and KNN Imputation for Handling Missing Data in Scholarship Applicant Datasets,” JUITA, vol. 13, no. 3, pp. 245–254, Nov. 2025, doi: 10.30595/juita.v13i3.26502.
[30] T. Mahmud, T. Akter, S. Anwar, M. T. Aziz, M. S. Hossain, and K. Andersson, “Predictive Modeling in Forex Trading: A Time Series Analysis Approach,” in 2024 Second International Conference on Inventive Computing and Informatics (ICICI), Bangalore, India: IEEE, Jun. 2024, pp. 390–397. doi: 10.1109/ICICI62254.2024.00070.
[31] K. Mahmud Sujon, R. Binti Hassan, Z. Tusnia Towshi, M. A. Othman, M. Abdus Samad, and K. Choi, “When to Use Standardization and Normalization: Empirical Evidence From Machine Learning Models and XAI,” IEEE Access, vol. 12, pp. 135300–135314, 2024, doi: 10.1109/ACCESS.2024.3462434.
[32] M. Waqas and U. W. Humphries, “A critical review of RNN and LSTM variants in hydrological time series predictions,” MethodsX, vol. 13, p. 102946, Dec. 2024, doi: 10.1016/j.mex.2024.102946.
[33] P. Nguyen-Duc, H. D. Nguyen, Q.-H. Nguyen, T. Phan-Van, and H. Pham-Thanh, “Application of Long Short-Term Memory (LSTM) Network for seasonal prediction of monthly rainfall across Vietnam,” Earth Sci Inform, vol. 17, no. 5, pp. 3925–3944, Oct. 2024, doi: 10.1007/s12145-024-01414-3.
[34] X. Wen and W. Li, “Time Series Prediction Based on LSTM-Attention-LSTM Model,” IEEE Access, vol. 11, pp. 48322–48331, 2023, doi: 10.1109/ACCESS.2023.3276628.
[35] R. Zhu, G.-Y. Zhong, and J.-C. Li, “Forecasting Price in A New Hybrid Neural Network Model with Machine Learning,” Expert Systems with Applications, vol. 249, p. 123697, Sep. 2024, doi: 10.1016/j.eswa.2024.123697.
[36] E. Seidi, F. Kaviari, and S. F. Miller, “Hyperparameter Tuning of Artificial Neural Network-Based Machine Learning to Optimize Number of Hidden Layers and Neurons in Metal Forming,” JMMP, vol. 9, no. 8, p. 260, Aug. 2025, doi: 10.3390/jmmp9080260.
[37] M. Jin and H. Emami-Meybodi, “Prediction of Flowing Bottomhole Pressure in Gas-Lifted Wells Using LSTM-ANN Models,” Energy Fuels, vol. 39, no. 31, pp. 14992–15002, Aug. 2025, doi: 10.1021/acs.energyfuels.5c02240.
[38] F. H. Sezgin, Ö. Algorabi, G. Sart, and M. Güler, “Hyperparameter-Optimized RNN, LSTM, and GRU Models for Airline Stock Price Prediction: A Comparative Study on THYAO and PGSUS,” Symmetry, vol. 17, no. 11, p. 1905, Nov. 2025, doi: 10.3390/sym17111905.
[39] I. Darmawan, D. P. Kemuning, A. Rahmatulloh, R. Rizal, E. Haerani, and L. Sutriani, “Optimizer Algorithms for Air Temperature Prediction using LSTM,” in 2025 9th International Conference on Recent Advances and Innovations in Engineering (ICRAIE), Kaula Lumpur, Malaysia: IEEE, Aug. 2025, pp. 1–6. doi: 10.1109/ICRAIE65839.2025.11239081.
[40] N. N. Karima, S. Ahmad, A. Islam, A. S. N. Huda, L. J. Awalin, S. A. Halim, H. Mokhlis, M. S. Ali, and S. Mubarok, “Enhancing Short-Term Load Forecasting Using Hyperparameter-Optimized Deep Learning Approaches,” Energies, vol. 19, no. 3, p. 705, Jan. 2026, doi: 10.3390/en19030705.
[41] H. Yüce and H. Arabacı, “Effect of Different Normalization Methods on Battery SoC Estimation Performance at the Optimal LSTM Batch Size,” in 2025 16th International Conference on Electrical and Electronics Engineering (ELECO), Istanbul, Turkiye: IEEE, Nov. 2025, pp. 1–5. doi: 10.1109/ELECO69582.2025.11329352.
[42] B. Liang, J. Deng, F. Yuan, J. Gao, J. Yang, and P. Zhao, “A bio-inspired IVY metaheuristic for optimizing LSTM models in urban traffic flow prediction,” Computing, vol. 108, no. 2, p. 29, Feb. 2026, doi: 10.1007/s00607-026-01623-2.
[43] P. Aduama, A. S. Al-Sumaiti, V. K. Saini, and R. Kong, “Explainable optimized voting ensemble for photovoltaic power forecasting,” Energy and AI, vol. 24, p. 100690, May 2026, doi: 10.1016/j.egyai.2026.100690.
[44] B. Wang, Z. Yang, M. Ji, J. Shan, M. Ni, Z. Hou, J. Cai, X. Gu, X. Yuan, Z. Gong, Q. Du, Y. Yin, and K. Jiao, “Long short-term memory deep learning model for predicting the dynamic performance of automotive PEMFC system,” Energy and AI, vol. 14, p. 100278, Oct. 2023, doi: 10.1016/j.egyai.2023.100278.
[45] A. Flores, H. Tito-Chura, J. Guzman-Valdivia, R. Morales-Gonzales, C. Rosado-Chavez, and C. Silva-Delgado, “Multi-Step Forecasting of Weekly Dengue Cases Using Data Augmentation and Convolutional Recurrent Neural Networks,” Computers, vol. 15, no. 2, p. 114, Feb. 2026, doi: 10.3390/computers15020114.
[46] D. Ramadhani, A. M. Soleh, and E. Erfiani, “Characteristics of Machine Learning-based Univariate Time Series Imputation Method,” JUITA, vol. 12, no. 2, p. 279–288, Nov. 2024, doi: 10.30595/juita.v12i2.23453.
[47] R. A. Putri, S. Surono, and A. Thobirin, “Evaluating LSTM Performance on Multivariate Time Series with One-Class SVM Outlier Detection,” JUITA, vol. 13, no. 2, pp. 127–134, Aug. 2025, doi: 10.30595/juita.v13i2.26160.
[48] R. R. Salim and D. Adytia, “Deep Learning-Based Sea Level Forecasting Using Informer in Cilacap, Indonesia,” JUITA, vol. 13, no. 3, pp. 317–327, Nov. 2025, doi: 10.30595/juita.v13i3.27176.
[49] S. Lang, M. Alexe, M. C. A. Clare, C. Roberts, R. Adewoyin, Z. Ben Bouallègue, M. Chantry, J. Dramsch, P. D. Dueben, S. Hahner, P. Maciel, A. Prieto-Nemesio, C. O’Brien, F. Pinault, J. Polster, B. Raoult, S. Tietsche, and M. Leutbecher, “AIFS-CRPS: ensemble forecasting using a model trained with a loss function based on the continuous ranked probability score,” npj Artif. Intell., vol. 2, no. 1, p. 18, Feb. 2026, doi: 10.1038/s44387-026-00073-7.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Dhia Rafifah Thifal, Aji Prasetya Wibawa, Adelia Desyana Eka Putri, Adelia Khansa Ristiaputri, Adhelia Wida Khaidir, Agung Bella Putra Utama
This work is licensed under a Creative Commons Attribution 4.0 International License.
JUITA: Jurnal Informatika is licensed under a Creative Commons Attribution 4.0 International License.
