This work focuses on developing and evaluating the efficacy of machine learning (ML) algorithms to estimate daily maize crop evapotranspiration under limited input weather variables in semi-arid climatic conditions. Five ML algorithms, namely, category boosting (CB), linear regression, support vector machine (SVM), artificial neural network (ANN), and stochastic gradient descent, were developed based on highly correlated inputs combinations and examined for ETc estimation at an experimental site, ICAR-IARI, New Delhi. The results show that CB- and ANN-based models were superior to other models in the training phase. However, the SVM- and ANN-based models were excellent during testing, regardless of the number of input variables used. Notably, the wind speed was the most influential in ETc estimation. By utilizing complete sets of weather inputs, the SVM- and ANN-based models produced promising outcomes, exhibiting coefficient of determination (R-2) of 0.986 and 0.987, mean absolute error of 0.121 and 0.124 (mm day(-1)), root-mean-square error of 0.172 and 0.165 (mm day(-1)), and mean absolute percentage error of 4.37 and 4.57%, respectively. Overall, these two models (i.e., SVM and ANN) were comparable; therefore, these models may be adopted for precise estimation of ETc in a semi-arid climate.