An improved wavelet de-noising-based back propagation neural network model to forecast the bioaerosol concentration

Abstract

Detection, monitoring and estimation of bioaerosol concentration have become increasingly important for several reasons, particularly for environmental quality monitoring and public health concerns on these materials. Bioaerosol concentration is highly variable and noisy, influenced by many factors including meteorological parameters. Directly predicting bioaerosol concentration with noisy data gives erroneous results. As such, we developed a facile approach to forecast bioaerosol concentration using the Wavelet De-noising-based Back Propagation (WDBP) neural network model. We used meteorological data for Changsha gathered from 1 November 2018 to 1 April 2019 to demonstrate the efficiency of WDBP neural network in forecasting atmospheric bioaerosol concentration. The superiority of the new approach over the single Back Propagation (BP) neural network was also validated using real dataset. Overall, the performance of WDBP neural network was satisfactory, underlying potential practical application of the method in forecasting bioaerosol concentration. This work provides a prospective model, useful in monitoring environmental quality and atmospheric bio-threats to public health.

Copyright © 2020 American Association for Aerosol Research

EDITOR:

• Nicole Riemer

Acknowledgments

We would like to express gratitude to Changsha Meteorological Bureau for providing the meteorological dataset used in this study. Special thanks to Huifenglong Economic and Technological Development Co. LTD for providing FDAI equipment. We also thank Huan Liu of Zhejiang University for her comments on this research. Dataset related to this study can be found at http://dx.doi.org/10.17632/bmtbd73nfm.1, an open-source online data repository hosted at Mendeley Data.

Disclosure statement

There is no conflict of interest.

Additional information

Funding

This research was privately funded by the researchers.