A review of machine learning for analysing accident reports in the construction industry
Bibliographic record
Abstract
Recently, there has been a growth in the research interest on applied machine learning (ML) in safety analysis in the construction industry. The increased interest is part of a search for improved prevention of occupational accidents with a focus on text analysis and natural language processing (NLP). However, ML-based approaches have been less adapted compared to their perceived benefits due to barriers of implementation and challenges in analysing safety records in the construction sector. And the current literature has been criticized for a lack of clarity around the description of methodologies, interpretation, and the context of the application. Therefore, this work aims to review the latest developments in research applying ML to accident report analysis in construction. A review of the published literature on ML-based analysis of construction accident reports was carried out and organized in terms of the data pre-processing, algorithms, testing and implementation and further organized based on data structure. The results of the review found limitation related to data availability besides the manual structuring and the less use of unsupervised learning reflect complexity of handling textual accident data. Moreover, types of accidents happen in proportionally varying frequencies and need careful tackling outside basic assumptions of data pre-processing in addition to the general need for data pre-processing comparative studies and automated pipelines. The review also showed that data mining (DM) and unsupervised learning were less used especially with semi-structured and unstructured datasets reflecting maybe inefficient natural language processing (NLP) application with these types of learning. Among the reviewed articles, only four out of six prototypes were externally validated on construction environment thus we propose that future efforts would benefit from incorporating a standardized development method that also explicit how ML safety recommendation informs decision making. Future research should experiment and ascertain different choices in the pre-processing stage, validating the performance of the ML models and implementation in the construction practices. Finally, there are more advanced NLP methods that could be applied if domain specific repositories were available such as relation extraction and there are various advances that could be explored including large language models (LLMs).
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 | 0.004 | 0.014 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.011 | 0.009 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
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".