The reliable prediction of corn yield for the United States of America is essential for effective food and energy management of the world. Three satellite-derived variables were selected, namely enhanced vegetation index (EVI), leaf area index (LAI) and land surface temperature (LST). The least absolute shrinkage and selection operator (LASSO) was used for regression, while random forest (RF), support vector regression (SVR) and long short-term memory (LSTM) methods were selected for machine learning. The three variables serve as inputs to these methods, and their efficacy in predicting corn yield was assessed in relation to evapotranspiration (ET). The results confirmed that a high level of performance can be achieved for yield prediction (mean predicted R2 = 0.63) by combining EVI + LAI + LST with the four methods. Among them, the best results were obtained by using LSTM (mean predicted R2 = 0.67). EVI and LST provided extra and unique information in peak and early growth stages for corn yield, respectively, and the usefulness of including LAI was not readily apparent across the whole season, which was consistent with the field growing conditions affecting the ET of corn. The satellite-derived data and the methods used in this study could be used for predicting the yields of other crops in different regions.