Application of a novel heterogeneous catalyst fluorite (calcined fluorspar CaF2) in biodiesel preparation from a highly viscous non-edible rubber seed oil has been studied. Optimization of process parameters and effect of process parameters on biodiesel conversion is done using response surface methodology (RSM) with central composite design analysis. The optimum process conditions, i.e., 12:1 molar ratio (mol/mol), 4 (wt%) catalyst, and 5 h of reaction time for a conversion of 95.95% oil to biodiesel conversion were observed. A quadratic regression model to predict the conversion in terms of process parameters was obtained. Artificial neural network (ANN) with the Levenberg-Marquardt algorithm is also attempted, with the best performance of 0.80239 at epoch-3 and a mean square error of 1.24. On comparison of the coefficient of determination, R-2, the value obtained from both RSM and ANN it is observed that the ANN model predicts conversion better than RSM, with minimum error.