Statistical Evaluation of the Performance of SVM Kernels for Air Quality Classification: A Case Study on India

Air pollution has now become a burning problem in the developing world, particularly in countries like India. Deteriorating air quality has led to a widespread increase in diseases related to the lungs in the population, leading to a stupendous increase in the economic burden on the general populace. Agriculture has also led to an increase in air pollution due to the stubble burning activity of the farmers. Every year in the months of winter, air pollution halts all the social and economic activities of major Indian metro cities. Air quality classification has emerged as one of the most significant research and modeling issues as a result of the significant increase in air pollution. Only if the data are properly classified will it be feasible to reduce the impact of air pollution on human health. Accurate air quality classification has become crucial for addressing these problems. However, classifying air quality data is challenging due to class imbalances. In this study, we tackle this issue by using the Support Vector Machine (SVM) approach with various kernels. Statistical evaluation showed that the linear kernel performed best with an accuracy of 0.78, followed closely by the radial kernel. These results suggest that SVM with a linear kernel can improve air quality classification, aiding in better preventive measures and emergency responses during severe pollution events.