Abstract
This publication revises the deteriorated performance of field calibrated low-cost sensor systems after spatial and temporal relocation, which is often reported for air quality monitoring devices that use machine learning models as part of their software to compensate for cross-sensitivities or interferences with environmental parameters. The cause of this relocation problem and its relationship to the chosen algorithm is elucidated using published experimental data in combination with techniques from data science. Thus, the origin is traced back to insufficient sampling of data that is used for calibration followed by the incorporation of bias into models. Biases often stem from non-representative data and are a common problem in machine learning, and more generally in artificial intelligence, and as such a rising concern. Finally, bias is believed to be partly reducible in this specific application by using balanced data sets generated in well-controlled laboratory experiments, although not trivial due to the need for infrastructure and professional competence.
Highlights
The effects of air pollution are well known and the health impact is massive, and with the increasing public awareness about the adverse health effects of air pollution, e.g., by fossil-fuel combustion, the urge to monitor and regulate the amount of hazardous gases or particulate matter (PM) is becoming even more important [1,2,3,4]
The situation is analogous for other machine learning (ML) algorithms such as neural network (NN) or random forest (RF); the reader is encouraged to consult fundamental statistics and ML literature for more details and explanations of all methods applied in this analysis [20,21,22,23]
Only detailed results for the NN model development are presented to demonstrate the added value of model inspection, and the RF and linear regression (LR) results are provided as supplementary material
Summary
The effects of air pollution are well known and the health impact is massive, and with the increasing public awareness about the adverse health effects of air pollution, e.g., by fossil-fuel combustion, the urge to monitor and regulate the amount of hazardous gases or particulate matter (PM) is becoming even more important [1,2,3,4]. More and more start-ups are entering the AQM market with novel low-cost sensor systems connected to the internet; some emerged from know-how in classical analytical chemistry, others from expertise in the Internet of Things [7,8]. The latter put too much trust in the used hardware, e.g., sensors, even though these often suffer from low performance due to interferences/cross-sensitivities, drifts, and large unit-to-unit variability, as a lot of research on low-cost sensors and devices has shown [9,10,11,12,13,14,15,16,17,18]. PM sensors are usually based on light scattering and the most significant interfering variable relates to water, as they appear to overestimate PM mass under high relative humidity [7,16]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
