Evapotranspiration (ET), which constitutes evaporation from soil and water surfaces and transpiration from stomata of plant leaves, is an important indicator for measuring global hydrological and carbon cycle balances. Though it is crucial to monitor ET for water resource management, energy production, and environmental conservation, predicting ET is a complex task and lacks a reliable approach for accurately predicting ET using remote sensing and meteorological data. ML methods, with their ability to handle complex and non-linear relationships to make accurate predictions, can be used to predict ET. In this study, ML algorithms—Random Forest Regression, Support Vector Regressor, Artificial Neural Network, and an ensemble model—are developed to predict forest evapotranspiration. The models are trained with ECMWF ERA5 reanalysis meteorological parameters (max. and min. air temperature, relative humidity, vapor pressure deficit, precipitation, volumetric soil water content, and wind speed), remote sensing data products (MODIS Enhanced Vegetation Index, MODIS Land Surface Temperature, MODIS Fractional Photosynthetically Active Radiation) as independent parameters, and ET data (8-day data) from MODIS as the target variable. All the datasets are interpolated to a 4-day temporal resolution for the years 2016–2018. From the ensemble model, a satisfactory R-squared value of 0.81 and RMSE value of 0.27 mm/day for the prediction were obtained using the parameters chosen from feature analysis. The trained model is used to predict the forest ET map for the Upper Aravali region for the years 2016–2018. Using ML algorithms to estimate ET rates can be useful for proactive resource management, particularly in water-stressed areas.