Given the role of gas turbines in electricity generation and the need for accurate monitoring of power plant performance, predicting generator output power under real operating conditions is of great importance. The main objective of this study is to develop a data-driven framework for predicting the output power of a gas turbine at full-load operation based on recorded environmental variables, including ambient temperature, ambient pressure, and relative humidity. Real power plant instrumentation data were used to train and test machine learning models after aggregation, preprocessing, removal of invalid records, identification of suspicious data points, and normalization. In this regard, the performance of artificial neural networks, support vector regression, random forest, and an ensemble random forest model was evaluated and compared using mean squared error, mean absolute error, root mean squared error, and the coefficient of determination. The results showed that the ensemble random forest model achieved the best overall performance by reducing prediction error and improving agreement with the measured values. The proposed framework can be used for intelligent monitoring of gas turbine output power, power plant generation planning, and supporting daily operational decision-making within the range of the available data.