2403.08013
SUPERVISED TIME SERIES CLASSIFICATION FOR ANOMALY DETECTION IN SUBSEA ENGINEERING
Ergys Çokaj, Halvor Snersrud Gustad, Andrea Leone, Per Thomas Moe, Lasse Moldestad
incompletemedium confidence
- Category
- Not specified
- Journal tier
- Specialist/Solid
- Processed
- Sep 28, 2025, 12:56 AM
- arXiv Links
- Abstract ↗PDF ↗
Audit review
The paper reports strong results but leaves key experimental details ambiguous (especially the unit of evaluation and the split protocol across correlated minute windows) and does not release the data, limiting independent verification. The candidate solution correctly identifies missing ingredients for faithful reproduction and flags the F1 discretization issue, but it overlooks that the paper does specify minute-level evaluation for the CNN and provides a concrete PCA setting (4 PCs) and CNN architecture sketch, so its critique is only partially accurate.
Referee report (LaTeX)
\textbf{Recommendation:} major revisions
\textbf{Journal Tier:} specialist/solid
\textbf{Justification:}
The paper presents a sensible comparative study with strong empirical outcomes, but it omits critical details required to establish the validity and reproducibility of those outcomes. In particular, the unit of evaluation and split protocol for classical pipelines are insufficiently specified and likely risk leakage if minute-level windows are split without grouping by series. The dataset is not publicly available, and seeds/splits are not provided, preventing independent verification. These issues warrant major revisions before firm conclusions can be drawn.