In the past decade, machine learning (ML) algorithms have been widely applied to build prediction models for various mining applications. However, no research has been reported that forecasts truck productivity using ML algorithms. In this study, two tree-based ensemble learning algorithms, including random forest (RF) and gradient boosting regression (GBR), were proposed in combination with Gaussian mixture modelling (GMM) to train prediction models of truck productivity. GMM was adopted as a clustering technique to extract a latent variable from the training dataset. Multiple linear regression (MLR) and decision tree (DT) as single learning algorithms were used to construct prediction models to be compared with the tree-based ensemble models. The results showed that the tree-based ensemble models performed better than single models in predicting truck productivity with and without GMM clustering. Furthermore, GMM significantly increased the predictability of truck productivity prediction models by considering the latent variable. From the relative importance analysis, haul distance was the most influential factor among the observed input variables. Finally, the GMM-RF and GMM-GBR models with high accuracy were the proposed models for predicting truck productivity at mine sites.