2019_Prediction Model of PM10 by Using Machine Learning
| Format: | General Document |
|---|
| _version_ | 1860797996628705280 |
|---|---|
| building | INTELEK Repository |
| collection | Online Access |
| collectionurl | https://intelek.unisza.edu.my/intelek/pages/search.php?search=!collection3 |
| copyright | Copyright©PWB2025 |
| country | Malaysia |
| date | 2019-10-06 |
| format | General Document |
| id | 15349 |
| institution | UniSZA |
| internalnotes | Sila masukkan subject wajib Dissertations, Academic. Terima kasih... |
| originalfilename | 15349_5772a4725c7980f.pdf |
| person | Nurul Latiffah Binti Abd. Rani |
| recordtype | oai_dc |
| resourceurl | https://intelek.unisza.edu.my/intelek/pages/view.php?ref=15349 |
| sourcemedia | Server storage Scanned document |
| spelling | 15349 https://intelek.unisza.edu.my/intelek/pages/view.php?ref=15349 https://intelek.unisza.edu.my/intelek/pages/search.php?search=!collection3 General Document Malaysia Library Staff (Top Management) Library Staff (Management) Library Staff (Support) Terengganu Faculty of Bio-resources & Food Industry English application/pdf 1.5 228 Server storage Scanned document Universiti Sultan Zainal Abidin UniSZA Private Access Universiti Sultan Zainal Abidin SAMBox 2.3.4; modified using iTextSharp™ 5.5.10 ©2000-2016 iText Group NV (AGPL-version) 15349_5772a4725c7980f.pdf 2019_Prediction Model of PM10 by Using Machine Learning Nurul Latiffah Binti Abd. Rani Air Quality – Mathematical Models 2019-10-06 Copyright©PWB2025 Particulate matter (PM10) is the key indicator of air pollution index (API) in Malaysia. Regardless of its importance in determining the API level in Malaysia, absence of some PM10 concentrations is noticed for certain day possibly because of equipment failure. Therefore, an attempt to develop an accurate PM10 prediction model was made. This study used parameters, such as meteorological (wind speed, wind direction, temperature, humidity) and air pollutants (NOx, NO, SO2, NO2, CO, O3) data obtained from 52 continuous air quality monitoring stations in Malaysia, beginning from January 2010 until December 2015. In this study, artificial neural network (ANN), multiple linear regression (MLR) models and input obtained from the principal component analysis (PCA) with high correlation (>0.75) were utilised to predict the PM10 concentration based on three classification groups by agglomerative hierarchical cluster (AHC), namely high polluted region (HPR), moderate polluted region (MPR) and low polluted region (LPR) in Malaysia. The input parameters obtained with excluded lagged PM10 concentration data for HPR, MPR, LPR through PCA were NOx, NO, NO2, CO, O3, temperature; NOx, NO, O3, wind direction and NOx, NO, NO2, temperature, respectively. The values of R2 and RMSE obtained from ANN by using these input parameters were 0.4736, 24.14; 0.1851, 23.63 and 0.0760, 19.66 for HPR, MPR and LPR, respectively. Meanwhile, the values of R2 and RMSE obtained from MLR by using these input parameters were 0.3407, 27.02; 0.1332, 23.98 and 0.0569, 20.05 for HPR, MPR and LPR, respectively. This analysis showed that the ANN model gave a better prediction as compared to the MLR. However, both models were less accurate by giving low R2 and high RMSE values among the regions. An attempt was made to enhance the model performance by including the lagged PM10 concentration data. The input parameters obtained with included lagged PM10 concentration data for HPR, MPR, LPR through PCA were NOx, NO, NO2, temperature, O3, PM10; NOx, NO, temperature, wind direction, PM10 and NOx, NO, NO2, temperature, PM10 , respectively. The values of R2 and RMSE obtained from ANN by using these input parameters were 0.7718, 16.11; 0.7590, 12.63 and 0.7721, 9.86 for HPR, MPR and LPR, respectively. Meanwhile, the values of R2 and RMSE obtained from MLR by using these input parameters were 0.7609, 16.27; 0.7442, 13.03 and 0.7642, 10.02 for HPR, MPR and LPR, respectively. Both models showed an accuracy enhancement when the lagged PM10 concentration data were included as one of the input data. By comparing the results, the ANN model still gave better prediction than the MLR for each region. In conclusion, ANN is a more accurate PM10 concentration prediction model as compared to MLR. However, both prediction models gave the best when lagged PM10 concentration data were applied as inputs, and gave higher and lower R2 and RMSE performance, respectively. Dissertations, Academic Sila masukkan subject wajib Dissertations, Academic. Terima kasih... RM10 Prediction Particulate Matter Machine Learning in Environmental Science Thesis |
| spellingShingle | 2019_Prediction Model of PM10 by Using Machine Learning |
| state | Terengganu |
| subject | Air Quality – Mathematical Models Dissertations, Academic |
| summary | Particulate matter (PM10) is the key indicator of air pollution index (API) in Malaysia. Regardless of its importance in determining the API level in Malaysia, absence of some PM10 concentrations is noticed for certain day possibly because of equipment failure. Therefore, an attempt to develop an accurate PM10 prediction model was made. This study used parameters, such as meteorological (wind speed, wind direction, temperature, humidity) and air pollutants (NOx, NO, SO2, NO2, CO, O3) data obtained from 52 continuous air quality monitoring stations in Malaysia, beginning from January 2010 until December 2015. In this study, artificial neural network (ANN), multiple linear regression (MLR) models and input obtained from the principal component analysis (PCA) with high correlation (>0.75) were utilised to predict the PM10 concentration based on three classification groups by agglomerative hierarchical cluster (AHC), namely high polluted region (HPR), moderate polluted region (MPR) and low polluted region (LPR) in Malaysia. The input parameters obtained with excluded lagged PM10 concentration data for HPR, MPR, LPR through PCA were NOx, NO, NO2, CO, O3, temperature; NOx, NO, O3, wind direction and NOx, NO, NO2, temperature, respectively. The values of R2 and RMSE obtained from ANN by using these input parameters were 0.4736, 24.14; 0.1851, 23.63 and 0.0760, 19.66 for HPR, MPR and LPR, respectively. Meanwhile, the values of R2 and RMSE obtained from MLR by using these input parameters were 0.3407, 27.02; 0.1332, 23.98 and 0.0569, 20.05 for HPR, MPR and LPR, respectively. This analysis showed that the ANN model gave a better prediction as compared to the MLR. However, both models were less accurate by giving low R2 and high RMSE values among the regions. An attempt was made to enhance the model performance by including the lagged PM10 concentration data. The input parameters obtained with included lagged PM10 concentration data for HPR, MPR, LPR through PCA were NOx, NO, NO2, temperature, O3, PM10; NOx, NO, temperature, wind direction, PM10 and NOx, NO, NO2, temperature, PM10 , respectively. The values of R2 and RMSE obtained from ANN by using these input parameters were 0.7718, 16.11; 0.7590, 12.63 and 0.7721, 9.86 for HPR, MPR and LPR, respectively. Meanwhile, the values of R2 and RMSE obtained from MLR by using these input parameters were 0.7609, 16.27; 0.7442, 13.03 and 0.7642, 10.02 for HPR, MPR and LPR, respectively. Both models showed an accuracy enhancement when the lagged PM10 concentration data were included as one of the input data. By comparing the results, the ANN model still gave better prediction than the MLR for each region. In conclusion, ANN is a more accurate PM10 concentration prediction model as compared to MLR. However, both prediction models gave the best when lagged PM10 concentration data were applied as inputs, and gave higher and lower R2 and RMSE performance, respectively. |
| title | 2019_Prediction Model of PM10 by Using Machine Learning |
| title_full | 2019_Prediction Model of PM10 by Using Machine Learning |
| title_fullStr | 2019_Prediction Model of PM10 by Using Machine Learning |
| title_full_unstemmed | 2019_Prediction Model of PM10 by Using Machine Learning |
| title_short | 2019_Prediction Model of PM10 by Using Machine Learning |
| title_sort | 2019_prediction model of pm10 by using machine learning |