A Novel Phase Space Reconstruction- (PSR-) Based Predictive Algorithm to Forecast Atmospheric Particulate Matter Concentration

Syed Ahsin Ali Shah,Wajid Aziz,Malik Sajjad Ahmed Nadeem,Majid Almaraashi,Seong-O. Shim,Turki M. Habeebullah

DOI: https://doi.org/10.1155/2019/6780379

2019-07-25

Scientific Programming

Abstract:The prediction of atmospheric particulate matter (APM) concentration is essential to reduce adverse effects on human health and to enforce emission restrictions. The dynamics of APM are inherently nonlinear and chaotic. Phase space reconstruction (PSR) is one of the widely used methods for chaotic time series analysis. The APM mass concentrations are an outcome of complex anthropogenic contributors evolving with time, which may operate on multiple time scales. Thus, the traditional single-variable PSR-based prediction algorithm in which data points of last embedding dimension are used as a target set may fail to account for multiple time scales inherent in APM concentrations. To address this issue, we propose a novel PSR-based scientific solution that accounts for the information contained at multiple time scales. Different machine learning algorithms are used to evaluate the performance of the proposed and traditional PSR techniques for predicting mass concentrations of particulate matter up to 2.5 micron (PM 2.5 ), up to 10 micron (PM 10.0 ), and ratio of PM 2.5 /PM 10.0 . Hourly time series data of PM 2.5 and PM 10.0 mass concentrations are collected from January 2014 to September 2015 at the Masfalah air quality monitoring station (couple of kilometers from the Holy Mosque in Makkah, Saudi Arabia). The performances of various learning algorithms are evaluated using RMSE and MAE. The results demonstrated that prediction error of all the machine learning techniques is smaller for the proposed PSR approach compared to traditional approach. For PM 2.5 , FFNN leads to best results (both RMSE and MAE 0.04 μ gm −3 ), followed by SVR-L (RMSE 0.01 μ gm −3 and MAE 0.09 μ gm −3 ) and RF (RMSE 1.27 μ gm −3 and MAE 0.86 μ gm −3 ). For PM 10.0 , SVR-L leads to best results (both RMSE and MAE 0.06 μ gm −3 ), followed by FFNN (RMSE 0.13 μ gm −3 and MAE 0.09 μ gm −3 ) and RF (RMSE 1.60 μ gm −3 and MAE 1.16 μ gm −3 ). For PM 2.5 /PM 10.0 , FFNN is the best and accurate method for prediction (0.001 for both RMSE and MAE), followed by RF (0.02 for both RMSE and MAE) and SVR-L (RMSE 0.05 μ gm −3 and MAE 0.04).

computer science, software engineering