Biodiesel has emerged as a promising renewable alternative to conventional diesel fuel, offering environmental advantages and supporting global sustainability efforts. This study applied machine learning techniques to predict biodiesel yield from rubber seed oil, a non-edible oil and locally available feedstock in Nigeria. A dataset of 20 experimental runs was used, with methanol-to-oil molar ratio, catalyst weight, temperature, and reaction time as input variables. Three models, Decision Tree Regressor (DTR), Random Forest Regressor (RFR), and Gradient Boosting Regressor (GBR) were developed and evaluated using R-squared and Mean Squared Error (MSE). The results obtained show that the square of the coefficient of regression (R2) for the DTR, RFR, and GBR were 0.7900, 0.8612, and 0.9937, respectively. The MSE for the DTR, RFR, and GBR were 13.64, 8.97, and 0.40, respectively. The Gradient Boosting Regressor performed best, showing the highest predictive accuracy for the biodiesel yield of 75.32 % at the optimum conditions of temperature (30 °C), time (75 min), catalyst loading (0.5g), methanol-to-oil ratio (4:1), and weight of methanol (10 g). The results revealed that the methanol-to-oil molar ratio had the most significant influence on biodiesel yield, with yields increasing as the ratio improved within optimal limits. The results demonstrate that machine learning can offer a cost-effective and time-saving alternative to labor-intensive experimental methods, thereby improving the ability to predict and optimize biodiesel production. The ensemble-based learning models, particularly GBR, have the potential to be reliable tools for biodiesel yield prediction and support their integration into renewable energy system development frameworks