Development of Fuel Consumption and CO2 Emission Prediction Models Based on Driving Behavior Analysis Using GA-Optimized Curve Fitting

Fuel consumption and CO2 emissions in road transport are strongly affected by how a vehicle is operated under real driving conditions. This study develops prediction models for fuel consumption and CO2 emissions using measurable vehicle-operation and driving-pattern indicators collected through On-Board Diagnostics (OBD) and smartphone-based tracking. The input variables include fuel flow rate per hour (FFRH), engine revolutions per minute (ERPM), speed (S), acceleration (A), and road grade (G). These variables are used as indirect indicators of driving patterns and operating conditions, rather than direct measurements of driver actions. A genetic algorithm (GA) is applied to optimize the coefficients of polynomial curve-fitting models and to obtain explicit mathematical equations for the two outputs. A dataset of 370 synchronized observations is used for model development. Several polynomial degrees and GA settings are examined to identify the most suitable model structure. The best fuel consumption model is obtained using a second-degree polynomial, while the CO2 emission model requires a third-degree polynomial. The results show that GA-optimized curve fitting can provide interpretable prediction equations with acceptable accuracy. The proposed approach may be useful for eco-driving analysis, traffic management, and preliminary assessment of emission-reduction strategies.