Air quality modeling for effective environmental management in Uttarakhand, India: A comparison of logistic regression and naive bayes

Introduction: Air pollution increases the load of hospitalization cases, especially for those who have respiratory problems. For effective environmental management, this study aims to compare the performance of two classification algorithms in machine learning (logistic regression and naive bayes) and to evaluate the selection of the best algorithm for predicting the air quality class. Materials and methods: Pollutants data (PM10, SO2, NO2) have been collected from the Haldwani, Kashipur and Rudrapur regions in Uttarakhand (India). In part I of the study, the Air Quality Index (AQI) is calculated and assigned a class accordingly. In part II, the performance of algorithms is compared, and the air quality class is predicted through the best algorithm. In part III, accuracy is calculated after comparing the predicted class with the actual class. Then, it is compared with the accuracy of our selected algorithm. Results: The study finds a positive correlation between PM10 and SO2 pollutants. The result shows that the highest accuracy is achieved through logistic regression to predict the air quality class. Further, logistic regression has achieved the same accuracy i.e., 98.70% after comparing predicted values with the actual values. Conclusion: Logistic regression is the best algorithm to predict the air quality class in the regions of Uttarakhand, where pollutants are being measured in the Government’s hospital. The research also indicates that asthma patients in the Kashipur and Rudrapur regions may experience more health effects due to moderately polluted air quality; however, the situation is improving during the monsoon season.