An enhanced machine learning-based rapid visual screening framework for low-rise RC buildings considering model uncertainty and decision threshold optimization
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Existing structures can become seismically vulnerable over time due to various factors including deterioration and outdated or seismic design provisions. Rapid visual screening (RVS) methods are commonly used to quickly filter large building inventories for at-risk structures, typically based on simple visual inspections, such as sidewalk surveys. In a previous study, the authors developed a machine learning (ML)-based RVS method for low-rise reinforced concrete (RC) buildings capable of identifying buildings that are likely to be severely damaged in an earthquake with an accuracy of 71%. However, uncertainty in the model’s predictions remains a concern. This study refines the previously proposed RVS methodology by addressing model uncertainty and minimizing misclassifications. Two primary approaches are proposed: the first analyzes class probabilities from the ML-based screening model to assess the prediction uncertainty rather than relying on the final predicted damage class. With this approach, buildings for which the ML model shows high uncertainty can be prioritized for more detailed evaluation. The second approach aims to optimize the decision threshold used by the ML model to more accurately identify buildings at risk of severe damage. This is done by evaluating the relative cost of misclassifications, low risk buildings identified as high risk (false positives) and high-risk buildings identified as low risk (false negatives). Building on the findings, this paper proposes a comprehensive three-level machine learning-based methodology for enhanced rapid seismic vulnerability assessments.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 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 it