Classifying ultrafine particle formation using machine learning on total number concentrations
Bibliographic record
Abstract
Ultrafine particles (UFP, less than 100 nm in aerodynamic diameter) pose significant health and environmental risks. These particles can originate from anthropogenic sources such as traffic emissions and fuel combustion, or form through atmospheric reactions. To reduce human exposure to UFP, it is important to understand the processes that can increase UFP concentrations. Current methods typically use particle size data to identify processes such as nucleation events. Recent studies have applied machine learning, particularly the transfer learning method, to this task. However, particle size distribution data is not widely available at many monitoring sites. In this study, we aimed to identify UFP sources using machine learning classification models only based on high time-resolution total number concentration data, rather than size distribution data, making the approach applicable to locations where only total UFP measurements are available. We have previously analyzed particle size data ranging from 6 to 520 nm collected from 2006 to the end of 2021 at the Southern Ontario Centre for Atmospheric Aerosol Research (SOCAAR) near a busy roadway in downtown Toronto, Canada. The days were classified into five categories: Strong Nucleation, Midday Pollution, Traffic Pollution, Baseline, and Mixed. Two machine learning approaches were tried: (1) ensemble learning models, and (2) transfer learning models. Both approaches performed well, achieving average accuracies of 80% and 75%, respectively. While transfer learning models were more robust to missing values, the ensemble learning method was slightly more robust to moderate noise and was also less computationally demanding. The ensemble learning models were then applied to a second monitoring location with a lower UFP level. The predicted classes at this site had similar characteristics to the classified days at the first site, with 78%, 58%, and 50% of the Strong Nucleation, Baseline, and Traffic Pollution days identified at both sites being classified into the same respective categories. These results suggest that the ensemble learning models could be transferred to other locations.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".