Carbon monoxide (CO) is the main component of vehicle exhaust in highway tunnels, which is one of the important parameters for the design of ventilation in road tunnels. It is very necessary to predict the CO concentration accurately and quickly to provide a reference for tunnel ventilation system control and mitigating the concentration of CO timely and to ensure the health and safety of the people in the tunnel. At the same time, it is beneficial to the green and energy-saving operation of the tunnel. Based on the field monitoring data of the road tunnel, Random Forest (RF) model of CO concentration prediction was established with the four influencing factors, i.e., monitor position, traffic, vehicle speed, wind speed. The model was employed to predict the concentration of CO in a 3 300-long road tunnel and the predictive values were compared with the real values. The results show that the CO concentration prediction model based on Random Forest has good prediction accuracy, which can predict the concentration of CO accurately. The RMSE and R2 of the testing dataset is 0.3979 and 0.9437, respectively, which is significantly better than those of the linear regression model and Support Vector Machine (SVM) model. The prediction model can be extended to predict the CO concentration of another tunnel. The RMSE and R2 of the validation dataset is 0.7276 and 0.7295, respectively. That is, the CO concentration in the tunnel can be predicted under the condition of known monitor position, traffic, vehicle speed, wind speed. The model provides a reference for tunnel ventilation system control or safety warning. The position of monitor point has the greatest influence on the CO concentration in the tunnel, and the CO gas concentration is the highest at the exit of the tunnel. With the increase of wind speed, the concentration of CO gas in the tunnel gradually decreases.