Multiple PM Low-Cost Sensors, Multiple Seasons’ Data, and Multiple Calibration Models

In this study, we combined state-of-the-art data analysis techniques (machine learning [ML] methods) and from low-cost particulate matter (PM) sensors (LCSs) to improve the accuracy of LCS-measured PM2.5 (PM with aerodynamic diameter less than 2.5 microns) mass concentrations. We collocated nine LCSs a reference instrument for 9 months, covering all local seasons, in Bengaluru, India. Using collocation data, evaluated performance trained around 170 ML models reduce observed bias PM2.5. The included (i) Decision Tree, (ii) Random Forest (RF), (iii) eXtreme Gradient Boosting, (iv) Support Vector Regression (SVR). A hold-out validation was performed assess model performance. Model metrics coefficient determination (R2), root mean square error (RMSE), normalised RMSE, absolute error. found that LCS measurements varied across different types (RMSE = 9–30 µg m-3) SVR best correcting measurements. Hyperparameter tuning improved (except RF). significant predictors (fewer number predictors, chosen based on recursive feature elimination algorithm) comparable ‘all predictors’ most better linear models. Finally, as research objective, introduced black carbon concentration into but no improvement