<p>Predicting river discharge under climate variability informs the design runoff considerations for the watershed. The application of artificial intelligence in modeling hydrologic events has aided the prediction of river discharge, especially under climate variability. However, the accuracy of the prediction depends on the selection of an appropriate and effective modeling tool. This study investigated the machine learning algorithms for prediction of river discharge under a climate change scenario in South Eastern Nigeria. Thirty years (1980–2010) of meteorological records of five rivers in the region, were used to construct, train, and validate various machine learning models to predict the discharge based on the climate variables (evaporation, sunshine, radiation, rainfall intensity, air temperature, wind speed, relative humidity, soil temperature, and atmospheric pressure). Statistical evaluation of the models showed that Ensemble Trees (ET) outperformed others. ET achieved the highest coefficient of determination (R²) with 66% for validation and 91% for testing. Gaussian Process Regression followed with 58% and 81%. Linear Regression, Support Vector Machine and Regression Trees recorded 60% and 64%, 63% and 70%, 58% and 85% respectively. Neural Network performed poorly in validation recording −&#xa0;69%. Further analysis using Compromise Programming Index (CPI) across all metrics confirms Ensemble Tree as the most reliable model for forecasting river discharge under climate change impact in the area. The result also revealed that in the study area, river discharge increases with increasing relative humidity. These findings are useful for water resources management in Southeastern Nigeria.</p>