Abstract Measuring the Research Octane Number (RON) and the Motor Octane Number (MON) at a low price is currently not feasible, thus making the use of predictive methods essential to accomplish this task. Nevertheless, the latter rely on expensive data and linear by volume models cannot be applied for complex fuels. In this work, we have investigated 41 parameters from inexpensive tests to find the inherent link between these fuel properties and the RON and the MON. To achieve this objective, we first reduced the number of properties to only consider the principal ones relying on principal component analysis (PCA). Then, we applied artificial neural network (ANN) to identify the underlying links between the properties and the rRON/MON. The measurement of the distillation curve, the atomic mass fraction and the specific gravity are the primary properties required for the current method. The achieved mean squared error (MSE) is equal to 0.7 [ON 2].