Determining additional oil recovery from a silica nano-fluid enhanced oil recovery project is essential prior to its field scale application to prepare economic feasibility of the project. Measurement of oil recovery is conventionally done through expensive and time-consuming laboratory methods. In pursuance of saving measurement’s cost and time, it is fundamental to develop fast and precise method, especially during the initial screening. This study uncovers the potential machine learning models, such as decision tree (DT) and extreme gradient boosting (XGBoost), as tools for predicting the expected oil recovery of silica-based nano-fluid flooding in sandstone reservoir. The application of mentioned tools for predicting the expected oil recovery has not been reported in the open literature. The models were developed using the results of 108 experiments. The model outputs were analyzed by performance matrices, such as root mean square error (RMSE), mean absolute error, mean square error, and co-efficient of determination (R$^2$). The input features were size of nano-particles, concentration of nano-particles, oil viscosity, oil density, salinity of water, and porosity and permeability of the rock. The results showed that both DT and XGBoost models are reliable to predict the oil recovery with high R$^2$ and low RMSE values of 0.9690 and 1.5837, and 0.9806 and 1.2608, for DT and XGBoost respectively. Heatmap analysis with Pearson's correlation co-efficient criterion showed that the porosity of rock, permeability of rock, concentration of nano-particles, and oil density have significantly more importance on the additional oil recovery.