Evaluation of nine machine learning regression algorithms for calibration of low-cost PM2.5 sensor

Kumar, Vikas; Sahu, Manoranjan

doi:10.1016/j.jaerosci.2021.105809

Cited by 55 publications

(27 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…RFR was a ML algorithm of choice in this study because of its high accuracy even with relatively small training datasets (Kumar and Sahu, 2021). However, if training of a large dataset is involved, other techniques such as XGBoost and neural networks could improve accuracy further than RFR.…”

Section: Discussionmentioning

confidence: 99%

“…Random Forest Regression (RFR) is an ensemble supervised ML algorithm used for a wide range of classification and regression predictive problems (Kumar and Sahu, 2021). Random forest involves constructing a large number of decision trees with each decision tree fitted on a different subset of the training dataset (also called Bagging), in addition to selecting a random subset of input variables at each split point in the construction of trees.…”

Section: Random Forest Regression Modelmentioning

confidence: 99%

“…we corrected for filter-based absorption measurements using Virkkula (2010) (referenced as "unrevised Virkkula" going forward), Virkkula equation with revised coefficients for the SGP site (referenced as "revised Virkkula"), the average of unrevised Virkkula and Ogren (2010) modified Bond (1999) correction (referenced as "Average"), and the Random Forest Regression (RFR), which is a supervised ensemble Machine Learning (ML) algorithm used for a wide range of classification and regression predictive problems (Kumar and Sahu, 2021). We provide an inter-comparison of the performances of these algorithms on the sampled SGP data.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Correcting for filter-based aerosol light absorption biases at ARM’s SGP site using Photoacoustic data and Machine Learning

Kumar¹,

Paik²,

Shetty³

et al. 2022

Preprint

View full text Add to dashboard Cite

Abstract. Measurement of light absorption of solar radiation by aerosols is vital for assessing direct aerosol radiative forcing, which affects local and global climate. Low-cost and easy-to-operate filter-based instruments, such as the Particle Soot Absorption Photometer (PSAP) that collect aerosols on a filter and measure light attenuation through the filter are widely used to infer aerosol light absorption. However, filter-based absorption measurements are subject to artifacts which are difficult to quantify. These artifacts are associated with the presence of the filter medium and the complex interactions between the filter fibers and accumulated aerosols. Various correction algorithms have been introduced to correct for the filter-based absorption coefficient measurements toward predicting the particle-phase absorption coefficient (Babs). Since previously-developed correction algorithms have a fixed analytical form, fundamentally, they are unable to predict particle phase absorption coefficients with a high degree of accuracy universally: different corrections are required for rural and urban sites across the world. In this study, we analyzed three months of high-resolution ambient data collected using a 3-wavelength photoacoustic spectrometer (PASS) and PSAP on the same inlet; both instruments were operated at the Department of Energy’s Atmospheric Radiation Measurement (ARM) Southern Great Plains user facility in Oklahoma. We implemented the two most commonly used analytical correction algorithms, namely the Virkkula (2010) and the average of Virkkula (2010) and Ogren (2010)-Bond (1999), as well as a Random Forest Regression (RFR) machine learning algorithm to infer particle phase Babs values from PSAP data and estimate their accuracy in comparsion to the refernce Babs values measured synchronously by the PASS. The wavelength averaged Root Mean Square Error (RMSE) values of Babs predicted using the RFR algorithm are improved by an order of magnitude in comparison to those predicted by the Virkulla (2010) and average correction algorithms. A revised form of the Virkkula (2010) algorithm suitable for the SGP site has been proposed; however, its performance yields approximately two fold errors when compared to the RFR algorithm. To further improve the accuracy of our proposed RFR algorithm, we trained and tested the model on dataset of laboratory-generated combustion aerosols. The RFR model used as inputs the size distribution, uncorrected Tricolor Absorption Photometer (TAP)-measured Babs, and nephelometer-measured scattering coefficient Bscat and predicted particle-phase Babs values within 5% of the reference Babs measured by a PASS. Machine learning approaches offers a promising path to correct for biases in long term filter-based absorption datasets and accurately quantify their variability and trends needed for robust radiative forcing determination.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Random Forest Regression Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Correcting for filter-based aerosol light absorption biases at ARM’s SGP site using Photoacoustic data and Machine Learning

Kumar¹,

Paik²,

Shetty³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…To eliminate the effect of overfitting problem and to identify a more appropriate learning algorithm, authors in ref. [45] evaluated the performance of 9 ML regression algorithms for calibration of low‐cost sensors which they were used for air‐pollution evaluation. The aim of this research work is also to investigate the performance of different ML regression algorithms for calibration of low‐cost PM sensors.…”

Section: Introductionmentioning

confidence: 99%

“…Although several ML approaches that have been used in this research work, overlaps with the work that has been done in ref. [45], however, this research further investigates two more additional ML algorithms such as Elastic Net and Stacking Regression in addition to the previously mentioned ML algorithms that have already been evaluated for calibration purposes and shown in ref. [45].…”

Section: Introductionmentioning

confidence: 99%

Evaluation and calibration of low‐cost off‐the‐shelf particulate matter sensors using machine learning techniques

Ghamari

Kamangir

Arezoo

et al. 2022

IET Wireless Sensor Systems

View full text Add to dashboard Cite

The use of inexpensive, lightweight, and portable particulate matter (PM) sensors is increasingly becoming popular in air quality monitoring applications. As an example, these low‐cost sensors can be used in surface or underground coal mines for monitoring of inhalable dust, and monitoring of inhalable particles in real‐time can be beneficial as it can possibly assist in preventing coal mine related respiratory diseases such as black lung disease. However, commercially available PM sensors are not inherently calibrated, and as a result, they have vague and unclear measurement accuracy. Therefore, they must initially be evaluated and compared with standardised instruments to be ready to be deployed in the fields. In this study, three different types of inexpensive, light‐scattering‐based widely available PM sensors (Shinyei PPD42NS, Sharp GP2Y1010AU0F, and Laser SEN0177) are evaluated and calibrated with reference instruments. PM sensors are compared with reference instruments in a controlled environment. The calibration is done by means of different machine learning techniques. The results demonstrate that the calibrated response obtained by fusion of sensors has a higher accuracy in comparison to the calibrated response of each individual sensor.

show abstract

Calibration of Low-Cost Air Sensors in the Presence of Pollutant Interferences: An Outdoor Evaluation

Souani,

Vigneron,

Maaref

et al. 2024

Lecture Notes in Civil Engineering

View full text Add to dashboard Cite

Evaluation of nine machine learning regression algorithms for calibration of low-cost PM2.5 sensor

Cited by 55 publications

References 37 publications

Correcting for filter-based aerosol light absorption biases at ARM’s SGP site using Photoacoustic data and Machine Learning

Correcting for filter-based aerosol light absorption biases at ARM’s SGP site using Photoacoustic data and Machine Learning

Evaluation and calibration of low‐cost off‐the‐shelf particulate matter sensors using machine learning techniques

Calibration of Low-Cost Air Sensors in the Presence of Pollutant Interferences: An Outdoor Evaluation

Contact Info

Product

Resources

About