Journal of New Approaches in Water Engineering and Environment
Scientific Journal

Comparison of Soft Computing-Based Artificial Intelligence Algorithms with the Transformer-Based TFT Model for Flood Modeling

Document Type : Original Article

Authors

Ramtin Tavoosi Rad 1
Mohammad Ansari ghojghar 1
Arash Malekian 2

1 Department of Reclamation of Arid and Mountainous Regions Engineering, Faculty of Natural Resources, University of Tehran, Karaj, Iran.

2 Dry and Mountainous Areas Restoration Engineering Group, Faculty of Natural Resources, University of Tehran

10.22034/nawee.2026.564165.1183
Abstract
Objective: This research, aiming to fill the existing research gap, evaluates the accuracy of three algorithms: (TFT) as a representative of the new generation of deep models based on Transformers, (GEP) as a symbolic evolutionary algorithm, and (ANFIS) as a soft computing model, in predicting daily peak discharge in the Taleghan watershed. Methods: Daily peak discharge data from five hydrometric stations over a 30-year period (1992-2021) were collected, refined, normalized, and transformed into four temporal input scenarios including the past 3, 6, 9, and 12 months. All models were trained in the R environment and evaluated using four indicators: NSE, MAE, RMSE, and R.
Results: The results indicated that the TFT model, by a significant margin, provided the best performance in all stations and scenarios. NSE values greater than 0.97 to 0.99, RMSE less than 0.1 and R close to 1 demonstrate the high capability of this model in accurately reconstructing discharge behavior. the GEP model had moderate performance but it performed more stably than ANFIS and showed relative superiority in some long-term horizons. The ANFIS model had the weakest performance.
Conclusions: TFT is by far the most accurate and reliable model for predicting daily peak discharge in the Taleghan watershed and is recommended for flood analysis on long-term time scales. GEP has acceptable performance but cannot compete with attention-based architectures. ANFIS, despite its efficiency in modeling non-linear relationships, has shown high sensitivity to temporal fluctuations in this study.

Keywords

GEP
ANFIS
Data Driven models
Flood managment
Ahmadian AhmadAbad,G., Zakeri Niri,M., 2021. Simulation and comparison of runoff estimation by neural network methods, regression and fuzzy inference(A case study:Dez catchment). Nivar, 45(114-115), 92-105(in persian). doi: 10.30467/nivar.2022.290750.1196.
Burrichter, B., Koltermann da Silva, J., Niemann, A., Quirmbach, M., 2024. A Temporal Fusion Transformer Model to Forecast Overflow from Sewer Manholes during Pluvial Flash Flood Events. Hydrology, 11(3), 41. https://doi.org/10.3390/hydrology11030041
Dehghani, N., Vafakhah, M., Bahremand, A., 2016. Rainfall-Runoff Modeling using Artificial Neural Network and Neuro-Fuzzy Inference System in Kasilian Watershed. J Watershed Manage Res. 7(13), 137-128(in persian).
Ebrahimi, A., Roozbahani, A., Bani habib, M.A., 2018. Groundwater Level Prediction Using Dynamic Bayesian Networks Based on Sensitivity Analysis (Case Study: Birjand Plain). Iranian Water, 12(2), 91-100. (In Persian)
Francisco, R., Matos, L.P., Marinheiro, R., Lopes, N., Portela, M.M., Barros, P., 2025. Application of Temporal Fusion Transformers to Run-Of-The-River Hydropower Scheduling. Hydrology, 12(81). https://doi.org/10.3390/hydrology12040081
Gauch, M., Lin, J., 2020. A Data Scientist’s Guide to Streamflow Prediction. arXiv:2006.12975v1
Jang, J. S. R., 1993. ANFIS: Adaptive-network-based fuzzy inference system. IEEE Transactions on Systems, Man, and Cybernetics, 23(3), 665–685.
Ji, W., Cao, Z., Li, X., 2023. Multi-Task Learning and Temporal-Fusion-Transformer-Based Forecasting of Building Power Consumption. Electronics, 12, 4656. https://doi.org/10.3390/ electronics12224656
Kavoosi,M., khozeymehnezhad,H., 2021. Review and compare performance of 4 modeling methods LS-SVM, NN, GEP and ANFIS-PSO in Simulation of Rainfall - Runoff (Study Area: Halil River - Jiroft Dam). Irrigation and Water Engineering, 11(3), 96-110. doi: 10.22125/iwe.2021.128115
Kontoni, D.-P. N., Onyelowe, K. C., Ebid, A. M., Jahangir, H., Rezazadeh Eidgahee, D., Soleymani, A., Ikpa, C., 2022. Gene Expression Programming (GEP) Modelling of Sustainable Building Materials including Mineral Admixtures for Novel Solutions. Mining, 2(4), 629-653. https://doi.org/10.3390/mining2040034
Lim, B., OArik, S., Loeff, N., Pfister, T., 2020. Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting. arXiv:1912.09363v3
Mastouri, R., Najarchi, M., Torabi, S.A., 2022. Application of group experimental mode decomposition algorithm to improve the accuracy of daily river flow prediction using artificial intelligence methods, Iranian Water Research Journal, 16(45), 1-14(in persian). magiran.com/p2469788
Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I: A discussion of principles. Journal of Hydrology, 10(3), 282–290. https://doi.org/10.1016/0022-1694(70)90255-6
Pornemat Roudsari, A., Ghaderi, K., Karimi Ghohari, SH., 2015. Rainfall Runoff Modeling using Group Method of Data Handling (GMDH) and Artificial Neural Network (ANN) IN In Polrood Basin. J Watershed Manage Res. 5(10), 68-84(in persian).
Rasiaya Koya, S., Roy, T., 2023. Temporal Fusion Transformers for Streamflow Prediction: Value of Combining Attention with Recurrence. https://doi.org/10.48550/arXiv.2305.12335
R.Koza, J., 1992. Genetic Programming as a Means for Programming Computers by Natural Selection. ISBN: 9780262527910
Salari Jozi, M., Ghorbani, KH., Sohrabian, E., Abdolhoseini, M., 2017. Daily River Flow Forecasting Using Data-Driven Models. Iranian Journal of Irrigation and Drainage, 4(10), 479-488(in persian).
Terzi, Ö., Özcanoğlu, O., 2017. STREAMFLOW ESTIMATION OF GÖKSU RIVER WITH GENE EXPRESSION PROGRAMMING. Journal of Engineering Sciences and Design, 5(3), 483-488. https://doi.org/10.21923/jesd.330479
Vakili, SH., Mousavi, M., 2022. Rainfall–runoff modeling using adaptive neuro-fuzzy inference system (ANFIS) and genetic algorithm (GA). Watter Supply, 20(10), 7460-7475. DOI:10.2166/ws.2022.318
Jamalvand, A., Yaghoubi, B., Izadbakh, M.A., Shabanlou, S., 2021. Optimization of ANFIS Network to Simulate Long Term Rainfall in Babolsar City. Iranian Journal of Soil and Water research, 52(8), 2111-2123(in persian). DOI:10.22059/ijswr.2021.323558.668970
Journal of New Approaches in Water Engineering and Environment
Volume 5, Issue 3
Summer 2026
Pages 73-93

Home

Guide for Authors

Submit Manuscript

Reviewers

Contact Us