Air Sense: Empowering Air Quality Prediction through Machine Learning Algorithms

Air quality has long been a paramount environmental concern for both the general public and society at large, with its adverse effects on human health being all too evident. The ability to predict air quality holds the promise of enabling behavioral adjustments across various levels, from individual choices to collective actions at the community, national, and even global scales. Predicting air quality, however, is a formidable challenge due to its substantial spatial variations, ranging from tranquil residential areas to bustling urban streets and industrial zones. Furthermore, atmospheric conditions, including factors like rainfall, air pressure, and temperature, exert a significant influence on pollutant levels in the air. Many existing air quality prediction models are hampered by issues like unbalanced data and the laborintensive nature of manual annotations. In this study, a novel approach to air quality prediction is introduced, harnessing the power of the Light GBM algorithm. The strategy involves rigorous feature selection through Recursive Feature Elimination (RFE), resulting in a remarkable performance by the LGBM model, as evidenced by significantly reduced error rates, including RMSE (9.78), MSE (95.60), MAPE (8.20), and MAE (8.45), coupled with an impressive R-Square value of 0.94. This underscores the model's exceptional capacity to elucidate the variance in AQI values, positioning it as a superior choice for air quality forecasting.