Driving patterns have a strong impact on fuel consumption and pollutant emission rates. Optimizing these factors can lead to more efficient and environmentally friendly driving. This study aims to categorize drivers into normal, timid, and aggressive types in order to estimate the differences in instantaneous fuel consumption. The data was obtained using a data acquisition device connected to the OBD port to extract the data identifier parameters of four vehicles monitored for 15 consecutive days, which belong to a service vehicle fleet. Through the implementation of a regression tree, we can estimate fuel consumption using the most important predictors. This method yields an accuracy of 99.7%, recall of 99.9%, and F1-score of 99.8%. Mathematical optimization is used to obtain the speed of the vehicle, accelerator pedal position, and engine speed that minimize fuel consumption for each gear. This study demonstrates that employing a normal driving style can lead to an average fuel flow of 3,015 \(\cdot 10^{-4}\) liters per second, which is significantly different from the fuel flow of 1,832 \(\cdot 10^{-3}\) liters per second associated with an aggressive driving style.