Comparison of Various Machine Learning Algorithms in Predicting Air Quality

Air pollution is the major environmental issue that need to be controlled immediately as it is causing various life-threatening disease like cancer, pneumonia, asthma, and other breathing problems to human beings. It is also damaging plants and soil and our ecosystem is getting imbalanced. Various human activities like burning of fossil fuels, factories and industries are becoming the major cause of air pollution. But before controlling it is necessary to predict air quality in terms of Air Quality Index. Various statistical methods exist to predict air quality but due to manual calculation it lacks accuracy. Machine learning plays an important role in predicting air quality accurately as it learns from training data. In our approach dataset is collecting from Open weather API. Preprocessing of dataset is done, and normalization of dataset is done using Z-score method. Time series prediction is followed with a delay of one hour. Filtration of dataset and statistical history for a particular date is also calculated. Visualization of dataset is also done in terms of graph. Input parameters taken are PM10. PM25, SO2, NO2, CO and output parameters are Air quality index. Prediction is done using K-Nearest Neighbor, Random Forest, and Naïve Bayes. Random forest is giving better performance as compared to other algorithms in terms of accuracy and Naïve Bayes exhibit least accuracy. Precision and Recall is also calculated for each category.