Estimation of Particulate Matter PM2.5 Concentration using Random Forest Regressor with Hyperparameter Tuning

Abstract

In recent years, study of particulate matter become an important public health concern. Small particals PM2.5, which have diameter less than 2.5 micro meter impacts on lung diseases and respiratory system of human. A number of various computational techniques are there to estimate the concentration of these particles present in the atmosphere. In this paper, Random Forest Regressor (RFR) is proposed to estimate the concentration of PM2.5 particles. Model is trained on 11 different features i.e. annual average temperature (T), maximum temperature (MT), minimum temperature (mT), rain precipitation (RP), average wind speed (WS), total rainy days (RD), total snowy days (SD), total stormy days (StD), total foggy days (FD), total tornado days (TD), total haily days (HD). Data is collected through web scrapping for the Bangalore city, India from year 2013 to 2020. Model performance obtained was R2=0.9732, MAE=3.87μg/m3, and RMSE=2.84μg/m3. Simulated result showed higher accuracy over other existing techniques.