A large portion of urban emissions in developing countries come from old gasoline vehicles driven in metropolitan areas. The present study aimed to develop models to estimate the environmental impact of different contents of gasoline and ethanol mixtures (pure gasoline; 25, 50, 75% ethanol blended to gasoline; and 100% ethanol) in a flex-fuel engine. We tested the blended fuel using three different speeds and recorded the GHG emissions and engine output data. The data mining approach was used to develop environmental impact predictive models. The ethanol content in gasoline; the engine rotational speed 900, 2000, and 3000 rpm; and λ were used as attributes. The classification target was the environmental impact concerning the CO2 emission ("low," "average," and "high"). We employed the Random forest algorithm to develop predictive models. The mean values of CO2 concentrations for all studied fuel content were above 2.47% of the volume. The trees' models (accuracy 73%, κ =0.61) showed three alternatives for predicting the environmental impact based on the ethanol blend, the engine rotation, λ, and the air-fuel ratio. Such models might help policymakers develop educational campaigns to reduce short- and medium-term urban commuter traffic pollution in countries that lack suitable urban transportation.