Comparative assessment of hybrid machine learning models for accurate groundwater level prediction in a data-limited aquifer

This dataset contains hydro-climatic and groundwater level (GWL) observations for the Astaneh-Kuchesfahan aquifer, Iran, covering a 256-month period (2002–2020). The data includes monthly records of GWL, precipitation, temperature, evapotranspiration, and water depth from two observation wells (C1 and C2). This dataset was compiled to evaluate the performance of various hybrid machine learning models—including ANN-SFOA, LSTM-TROA, ANFIS-EGOA, RF-Wavelet, and GAN-RF—in predicting groundwater levels. It is particularly valuable for research focused on time-series analysis, hydrogeological modeling, and the application of machine learning in water resource management. By providing multi-scale patterns, this dataset enables the study of predictive model generalization and resistance to overfitting. It supports the research findings regarding the superiority of the RF-Wavelet model in denoising and enhancing predictive accuracy. Users are encouraged to cite the associated publication when utilizing this data for further scientific investigations.本数据集包含伊朗阿斯塔内-库切斯法汉含水层（Astaneh-Kuchesfahan aquifer）的水文气候与地下水位（groundwater level, GWL）观测数据，时间跨度为256个月（2002年至2020年）。数据涵盖两口观测井（C1与C2）的地下水位、降水量、气温、蒸散发以及水深的月度记录。本数据集旨在用于评估各类混合机器学习模型（包括ANN-SFOA、LSTM-TROA、ANFIS-EGOA、RF-Wavelet以及GAN-RF）在地下水位预测任务中的表现。该数据集对于时间序列分析、水文地质建模以及机器学习在水资源管理中的应用相关研究具有重要价值。通过提供多尺度数据模式，本数据集可支撑预测模型泛化能力与抗过拟合性能的相关研究。本数据集可佐证RF-Wavelet模型在去噪与提升预测精度方面的性能优势相关研究结论。若使用者基于本数据集开展后续科学研究，敬请引用相关发表文献。