Evaporation from dam reservoirs is a major source of water loss in arid and semi-arid regions, making accurate prediction essential for sustainable water resource management. In this study, evaporation from the Amir Kabir Dam reservoir was predicted for 2020–2024 using statistical and machine learning approaches. Monthly evaporation was selected as the target variable, while month, year, air temperature, solar radiation, relative humidity, and wind speed were used as input parameters. Three models—Multivariate Linear Regression (MLR), Random Forest (RF), and XGBoost—were evaluated using mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²). Correlation analysis and feature importance assessment were also performed to identify the dominant factors affecting evaporation. The results showed that the machine learning models outperformed the linear regression model in predicting evaporation dynamics. Among the tested models, RF achieved the highest accuracy (MAE = 115.35, RMSE = 156.02, R² = 0.8755), followed by XGBoost (R² = 0.8651), while MLR showed the lowest performance (R² = 0.5157). The findings confirmed the nonlinear and multifactorial nature of reservoir evaporation. Correlation analysis indicated positive relationships between evaporation and solar radiation, wind speed, and air temperature, whereas relative humidity showed a negative correlation. Feature importance analysis identified solar radiation as the most influential variable in both RF and XGBoost models, while the month variable highlighted the role of seasonal variations in evaporation behavior. Overall, the results demonstrate that machine learning approaches, particularly RF, are effective tools for reservoir evaporation prediction and can support water resource management and monitoring of high-risk evaporative conditions.