The production of wild blueberries (Vaccinium angustifolium.) is contributing 112.2 million dollars to the Canada’s revenue which can be further increased through controlling harvest losses. A precise prediction of blueberry harvesting losses is necessary to mitigate such losses. In this study, the performance of three machine learning (ML) models was evaluated to predict the wild blueberry harvest losses on the ground. The data from four commercial fields in Atlantic Canada were used for this purpose. Wild blueberry losses (fruit loss on ground, leaf losses, blower losses) and yield was measured manually from randomly selected plots during mechanical harvesting. Wild blueberry plant height, fruit zone, and field slope readings were recorded from each of the plots. Three ML models namely linear regression (LR), support vector regression (SVR), and random forest (RF) were used to predict ground losses as a function of plant height, fruit zone, slope, fruit yield, leaf loss, and blower loss. Statistical parameters i.e., root mean square error, mean absolute error, and coefficient of determination (R²) were used to assess the prediction accuracy of the models. Correlation analysis revealed that fruit yield and other losses (leaf loss, blower loss) had moderate to high correlations judged from the coefficient of correlation (r), i.e., r = 0.37- 0.79. The LR model had the best predictions for ground losses among all the tested models. Frank Webb, Tracadie, Cooper, and Small Scott fields had a R² values of 0.91, 0.87, 0.73, 0.91, respectively. SVR performed better for Frank Webb (R² = 0.93), Tracadie (R² = 0.88) and Cooper (R² = 0.79) except for Small Scott (R² = 0.07). In the comparison of actual and predicted ground losses, the SVR outperformed (R² = 0.79-0.93) the other two models followed by LR (R² = 0.73 to 0.92) for three fields. The results revealed that these ML models could be useful in the prediction of ground losses during the harvesting of wild blueberries in the selected fields.