There is a serious health threat posed by air pollution and the study of this paper evaluates the accuracy of machine learning in determining the concentration of tropospheric ozone that exists in cities found in India. It comprises three predictive models, including the Random Forest, AdaBoost, and Support Vector Machines (SVM) having real-time seven-day prediction window. The models utilize statistical scaling factors and the meteorological conditions and the number of pollutants at more than 300 government-controlled monitoring stations in over 100 cities in India through the information provided by the Central Pollution Control Board (CPCB). The data set exhibits seven key atmospheric pollutants such as PM 2.5, PM 10, NO 2, SO 2, CO, NH 3, and O 3. When all the tested algorithms were compared with one another, Random Forest showed better predictive performance with a result of 0.79 in the R2 value and a root-mean-square error (RMSE) of 12.68 0 g/m3, and a mean absolute error (MAE) of 8.94 0 g/m3. The analysis of feature importance by the use of the Random Forest model revealed that the pre-day ozone levels, maximum temperature, and NO 2 concentrations are the most significant variables. Also, the predictive accuracy of the model had an inverse relationship with the concentration of ozone such that the model showed a better performance when the concentration was low to moderate but not very high exceedances. The outcomes are the perfect foundation to transfer machine learning to the actions of the public health and enable the authorities to warn people about the state of the air quality in time and employ measures to regulate the pollution. This study observes the possibility of using real-time environmental information to make short-term ozone predictions in aid of air quality management in urban environs.