The control of GDI (gasoline direct injection) engine exhaust emissions especially NOx and soot particles is the major challenging task because direct injection of fuel with high pressure leads to wall impingement of fuel and higher in-cylinder temperature with improved combustion. In this study, the CNG fueled vehicle emissions were predicted with the help of the NN model (Neural Network) of 3WCC (3-Way Catalytic Converter) using AMESim simulation tool. The framework replaced the complex 3WCC physical model into a NN based reduced order model. The NN model receives the input parameters such as vehicle velocity (or) acceleration, engine torque, engine speed, fuel consumption and catalyst inlet temperature; and predicts post catalyst emission mass flows and catalyst outlet temperature. The NN model was trained using real time WLTC (Worldwide Harmonized Light-duty Test Cycle). The NN model minimizes the computational load, captures dynamic catalyst behaviour and allows precise control of emissions during transient driving cycles. The exhaust CO2 cumulative mass exactly follows the same path of baseline, whereas the HC and CO cumulative masses of the NN model showed greater prediction than the baseline. The greater prediction of CO results in the same pattern of soot particle emission because of the rich mixture is the reason for both CO and soot particle emissions. The NOx cumulative mass of the NN model showed a similar pattern to that of baseline with slight gradual increment during transient conditions. At the end of the cycle, the NN model showed the reduction of CO, HC and NOx emissions around 15.38%, 20% and 15% respectively compared with the baseline emission concentrations. Overall, the well trained NN reduced order model showed superior accuracy towards minimizing the engine exhaust emission concentrations.