Application of Step Wise Regression Analysis in Predicting Future Particulate Matter Concentration Episode

Abstract

Particulate Matter is an air pollutant that has resulted in tremendous health effects to the exposed populace. Air quality forecasting is an established process where air pollutants particularly, Particulate Matter (PM10) concentration is predicted in advance, so that adequate measures are implemented to reduce the health effect of PM10 to the barest level. The present study used daily average PM10 concentration and meteorological parameters (temperature, humidity, wind speed and wind direction) for five years (2006)(2007)(2008)(2009)(2010) from three industrial air quality monitoring stations in Malaysia (Balok Baru, Tasek and Paka). Time series plot was used to assess PM10 pollution trend in the industrial areas. Additionally, Step Wise Regression (SWR) analysis was used to predict next day PM10 concentrations for the three industrial areas. The SWR method was compared with a Persistence model to assess its predictive capabilities. The results for the trend analysis showed that, Balok Baru (BB) had higher PM10 concentration levels, having high values in 2006, 2007 and 2009. These values were higher than the Malaysian Ambient Air Quality Guideline (MAAQG) of 150 µg/m 3 . Subsequently, the other two industrial areas Tasek (TK) and Paka (PK) had no record of violating the MAAQG. The results for the SWR analysis had significant R 2 values of 0.64, 0.66 and 0.60, respectively. The model performance results for Variance Inflation Factor (VIF) were less than 5 and Durbin Watson test (DW) had value of 2 for each of the study areas, which were significant. The comparative analysis between SWR and Persistence model showed that the SWR had better capabilities, having lower errors for the BB, TK and PK areas. Using Root Mean Square Error (RMSE), the results showed error differences of 7, 12 and 16%, and higher predictability using Index of Agreement (IA), having a difference of 17, 19 and 16% for BB, TK and PK areas, respectively. The results showed that SWR can be used in predicting PM10 next day average concentration, while the extreme event detection results showed that 100 mg/m 3 were better detected than the 150 mg/m 3 bench marked levels.

Application of Artificial Neural Networks in the Prediction of PM10 Levels in the Winter Months: A Case Study in the Tricity Agglomeration, Poland

“…Researchers have used statistical modelling techniques and machine learning methods to analyse, engage the proper variables within modelling framework, and, finally, predict the concentrations of particulate matter. The most preferred approaches are multiple linear regression, stepwise regression, artificial neural networks, principal component analysis, and clustering methods [21][22][23][24][25][26][27][28][29]. It should be noted that a number of studies have compared the performance of various modelling approaches to determine the best model for the prediction of PM 10 in different locations.…”

Documenting the Constraints and Its Allied Factors in Agar (Aquilaria malaccensis Roxb.) Tree Cultivation at Farm Level: A Case Study in Bangladesh

“…The factors related to agar farmers' restrictions in agar production were determined using multiple regression analysis (entry technique) [25]. The multiple regression analysis equation is as follows (3):…”

Investigation of PM10 prediction utilizing data mining techniques: Analyze by topic

“…The stepwise regression was a combination of the forward selection and the backward elimination. It was implemented by a number of researchers (Biancofiore et al, 2017;Chaloulakou et al, 2003;Díaz-Robles et al, 2008;Grivas & Chaloulakou, 2006;Huebnerova & Michalek, 2014;Kim Oanh & Leelasakultum, 2011;Nazif et al, 2016Nazif et al, , 2018Russo et al, 2015;Stadlober et al, 2008;Vlachogianni et al, 2011;Vlachogiannis & Sfetsos, 2006).…”