Froth flotation is a multiphase process which exhibits inherent instability and complex dynamics. As a result of this, advances in predicting the overall performance of the process is of much significance to mineral processing engineers. In this work, we present a comparative rougher copper recovery prediction study using four predictive algorithms ie support vector machine, Gaussian process regression (GPR), artificial neural network and linear regression. Each predictive algorithm was trained and validated with 80 per cent and 20 per cent of the total data set received, respectively. Additionally, the various trained models were further assessed with an independent data set (2000 observations) collected on a later date than the training and validation data sets. Models performance assessment using correlation coefficient (𝑟𝑟), root mean square error (RMSE), mean absolute percentage error (MAPE) and variance accounted for (VAF) indicated that GPR model performs better than all other models attaining 𝑟𝑟 values >0.96, RMSE and MAPE values <0.42 and><0.25 per cent, respectively, and vaf values>94 per cent using training, validation and testing set. The overall performance of the GPR model indicates that good rougher copper recovery prediction can be made using GPR model and relevant rougher flotation variables.