Evaluation of white-box versus black-box machine learning models in estimating ambient black carbon concentration

Fung, Pak Lun; Zaidan, Martha Arbayani; Timonen, Hilkka; Niemi, Jarkko V.; Kousa, Anu; Kuula, Joel; Luoma, Krista; Tarkoma, Sasu; Petäj̈ä, Tuukka; Kulmala, Markku; Hussein, Tareq

doi:10.1016/j.jaerosci.2020.105694

Cited by 39 publications

(49 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The penetration efficiency was ~47% for 0.01 μm, ~93% for 0.3 μm and ~40% for 10 μm (Hussein et al, 2020). These deficiency of measurement at the upper and lower edges is somewhat in alignment with other literatures.…”

Section: Measurement Sites and Instrumentssupporting

confidence: 79%

“…In warmer months, the diurnal cycles are not as distinct, but a sharp peak of nucleation mode around noon is found, which is associated with the occurrence of new particle formation. These events occurred very often in the summer as suggested by Hussein et al (2020). The amplitude of diurnal cycles of coarse mode is small while the patterns of accumulation are not clear ( Figure 6).…”

Section: General Pattern Of Particle Size Distributionmentioning

confidence: 84%

“…The meteorological measurement was performed with a weather station (WH-1080, Clas Ohlson: Art.no.36-3242, Helsinki, Finland) with a time resolution of 5 minutes. The meteorological data were comprised of ambient temperature (Temp, resolution 0.1℃), relative humidity (RH, resolution 1%), wind speed (WS), wind direction (WD, 16 equal divisions) and air pressure (P, resolution 0.3 hPa) (Hussein et al, 2019;Hussein et al, 2020;Zaidan et al, 2020). Wind direction is resolved into north and east direction, as WD-N and WD-E, respectively.…”

Section: Measurement Sites and Instrumentsmentioning

confidence: 99%

“…After data collection and data pre-processing procedures, the next step is model optimisation (Figure 1). ANN models have been utilised in predicting air quality (Freeman et al, 2018;Maleki et al, 2019;Cabaneros et al, 2019;Zaidan et al, 2020). Neural networks provide a robust approach for approximating real-valued target functions because they can mimic the non-linearity of the functions and their optimization methods are well developed (Zaidan et al, 2017).…”

Section: Modellingmentioning

confidence: 99%

“…Another data-driven approach black-box (BB) models, which refer to ML systems being viewed in terms of its inputs and outputs, without any knowledge of its internal workings or underlying principles (Rudin, 2019). They are considered to work generally better in terms of accuracy, but provide limited transparency and accountability regarding the results (Zaidan et al, 2019;Fung et al, 2020). One example of BB model is artificial neural network (ANN), which were applied extensively to estimate other air pollutant parameters (Freeman et al, 2018;Cabaneros et al, 2019).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Neural network modelling to estimate particle size distribution based on other particle sections and meteorological parameters

Fung¹,

Zaidan²,

Surakhi³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. In air quality research, often only particle mass concentrations as indicators of aerosol particles are considered. However, the mass concentrations do not provide sufficient information to convey the full story of fractionated size distribution, which are able to deposit differently on respiratory system and cause various harm. Aerosol size distribution measurements rely on a variety of techniques to classify the aerosol size and measure the size distribution. From the raw data the ambient size distribution is determined utilising a suite of inversion algorithms. However, the inversion problem is quite often ill-posed and challenging to invert. Due to the instrumental insufficiency and inversion limitations, models for fractionated particle size distribution are of great significance to fill the missing gaps or negative values. The study at hand involves a merged particle size distribution, from a scanning mobility particle sizer (NanoSMPS) and an optical particle sizer (OPS) covering the aerosol size distributions from 0.01 to 0.42 μm (electrical mobility equivalent size) and 0.3 μm to 10 μm (optical equivalent size) and meteorological parameters collected at an urban background region in Amman, Jordan in the period of 1st Aug 2016–31st July 2017. We develop and evaluate feed-forward neural network (FFNN) models to estimate number concentrations at particular size bin with (1) meteorological parameters, (2) number concentration at other size bins, and (3) both of the above as input variables. Two layers with 10–15 neurons are found to be the optimal option. Lower model performance is observed at the lower edge (0.01

show abstract

Section: Measurement Sites and Instrumentssupporting

confidence: 79%

Section: General Pattern Of Particle Size Distributionmentioning

confidence: 84%