Predictive Model with Machine Learning for Environmental Variables and PM2.5 in Huachac, Junín, Perú

Atmosphere, Journal Year: 2025, Volume and Issue: 16(3), P. 323 - 323

Published: March 12, 2025

PM2.5 pollution is increasing, causing health problems. The objective of this study was to model the behavior PM2.5AQI (air quality index) using machine learning (ML) predictive models linear regression, lasso, ridge, and elastic net. A total 16,543 records from Huachac, Junin area in Peru were used with regressors humidity % temperature °C. focus environmental variables. Methods: Exploratory data analysis (EDA) applied. Results: has high values winter spring, averages 52.6 36.9, respectively, low summer, a maximum value September (spring) minimum February (summer). use regression produced precise metrics choose best for prediction PM2.5AQI. Comparison other research highlights robustness chosen ML models, underlining potential Conclusions: found α = 0.1111111 Lambda λ 0.150025, represented by 83.0846522 − 10.302222000 (Humidity) 0.1268124 (Temperature). an adjusted R2 0.1483206 RMSE 25.36203, it allows decision making care environment.

Language: Английский

Exploring PM2.5 and PM10 ML forecasting models: a comparative study in the UAE

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 21, 2025

Particulate Matters PM $$_{2.5}$$ and $$_{10}$$ present a major health environmental concern in urban regions. This research compares machine learning time series models, such as Decision Tree (DT), Random Forest (RF), Support Vector Regression (SVR), Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), Facebook Prophet, for predictions of these matters. Their performances have been evaluated over 1-2 hours, 1 day week forecasting periods using five years real-life data from six ground stations Abu Dhabi, UAE. Performance metrics including Mean Absolute Percentage Error (MAPE), Root Squared (RMSE), (MAE), Percent Bias (PBIAS) were applied. Linear SVR was generally the best performing model at all with averages 18.7% 28.2% MAPE 2-hour periods, respectively. However, CNN performed 1-hour horizon, an average 12.6%. For forecast, outperformed other 18.3% MAPE. Prophet consistently others both 21.8% 13.4% 1-day 21.3% 13.8% 1-week, These models yielded similar RMSE, MAE, PBIAS values .

Language: Английский

Application of Deep Learning Techniques for Air Quality Prediction: A Case Study in Macau

Processes, Journal Year: 2025, Volume and Issue: 13(5), P. 1507 - 1507

Published: May 14, 2025

To better inform the public about ambient air quality and associated health risks prevent cardiovascular chronic respiratory diseases in Macau, local government authorities apply Air Quality Index (AQI) for management within its jurisdiction. The application of AQI requires first determining sub-indices several pollutants, including respirable suspended particulates (PM10), fine (PM2.5), nitrogen dioxide (NO2), ozone (O3), sulfur (SO2), carbon monoxide (CO). Accurate prediction is crucial providing early warnings to before pollution episodes occur. improve accuracy, deep learning methods such as artificial neural networks (ANNs) long short-term memory (LSTM) models were applied forecast six pollutants commonly found AQI. data this study was accessed from Macau High-Density Residential Monitoring Station (AQMS), which located an area with high traffic population density near a 24 h land border-crossing facility connecting Zhuhai Macau. novelty work lies potential enhance operational forecasting ANN LSTM run five times, average pollutant forecasts obtained each model. Results demonstrated that both accurately predicted concentrations upcoming h, PM10 CO showing highest predictive reflected Pearson Correlation Coefficient (PCC) between 0.84 0.87 Kendall’s Tau (KTC) 0.66 0.70 values low Mean Bias (MB) 0.06 0.10, Fractional (MFB) 0.09 0.11, Root Square Error (RMSE) 0.14 0.21, Absolute (MAE) 0.11 0.17. Overall, model consistently delivered PCC (0.87) KTC (0.70) lowest MB (0.06), MFB (0.09), RMSE (0.14), MAE (0.11) across all SD (0.01), indicating greater precision reliability. As result, concludes outperforms offering more accurate consistent tool management.

Language: Английский