The use of personnel cars has increased recently, which has clearly resulted in a reduction of quality of the air. Pollutant emissions from moving cars are the main cause of air pollution in metropolitan cities. One of those main pollutants released creating a negative impact on the environment is carbon monoxide (CO). The current study was done at three different four-legged signalized intersections in the metropolitan city of Hyderabad to investigate how several aspects of traffic stream, such as approach volume (ATV), average queue length during red-light wait time (QL), area of intersection (AOI), and cycle time length (CT), affect CO concentrations (COC) levels at signalized intersections (SI). When predicting CO at SI, two models employing multiple linear regression (MLR) and support vector machines (SVM) have been considered. CO levels at three signalized junctions were obtained as 9 ppm for Gandimaisamma crossroads and 12 ppm for both Bachupally and Pragathinagar HTL crossroads. The increment in COC with increase of CT, QL and ATV and decrease of AOI was noted due to the rapid movements of the vehicles at the signalized intersections. The models exhibit a certain simulated COC and field COC are resulting same sets of roadway and traffic conditions through validation. The results shown that MLR having a prediction accuracy of 0.96 with R2, MAPE of 4.4, whereas R2 obtained using SVM is 0.98 and MAPE of 4.04, which indicates SVM based simulation predicts the CO concentration more accurately than MLR simulation.