Emissions from vehicles at signalized intersections are one of the major sources of urban pollution, more prevalent in mid-size cities of India due to heterogeneous traffic conditions. The current study analyzes concentration of six pollutants (CO2, TVOC, PM2.5, CO, PM10 and HCHO) emitted at intersections of Warangal, Tirupati and Vijayawada. Information was collected on traffic volume, signalization duration, traffic queues and concentration of the mentioned pollutants in several intersections of those three cities. MLR, SVR and ANN models were built based on the collected information as a function of traffic volume and signal parameters. High correlation was detected between the concentration of all the pollutants and the studied parameters, particularly with approach volume and average length of queue. The ANN model provided the highest prediction accuracy among the tested models, with R² higher than 0.85 in all cases. Diurnal patterns indicate maximum emissions occurring in evenings corresponding to maximum traffic flows. Validation of the models proved statistical reliability of the models that can be applied in practice.