REVIEW 3 major objections 5 minor 16 references
Predicting Estimated Times of Restoration for Electrical Outages Using Longitudinal Tabular Transformers
T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash
Pith's one-line read A model that reads the entire sequence of outage updates predicts restoration times and improves a customer-satisfaction score by an average of 19.08%.
desk verdict Sensible longitudinal framing and useful customer-centered metrics, but the headline CSI gain is not credible yet because the baselines get less information and the evaluation metric equals the training loss. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the Longitudinal Tabular Transformer: each outage event is a variable-length sequence of updates; categorical features are embedded, concatenated with Z-normalized continuous features, linearly projected to $d_{\text{model}}=128$, and added to a positional encoding of the time delta $\Delta t_{i,j}$ since the first update. Six transformer encoder layers with 16 heads, masked for variable length, attend over the update sequence, and the representation of the last update is passed through a fully connected head to predict the ETR. The second piece of machinery is the asymmetric scoring scheme: a piecewise loss $\ell(\epsilon_i)$ with penalties $\alpha=5$ for under-prediction, $1$ for over-prediction within eight hours, and $\beta=2$ for over-prediction beyond eight hours, together with the derived metrics UPR, OPR-8, WAE, and CSI.
What would settle it
Retrain every baseline on the same data the LTT sees—for instance, by flattening all updates into one feature set or feeding the update sequence to a standard sequence model—and compare the Customer Satisfaction Impact. If the 19.08% average improvement shrinks to a few points or disappears, the paper's claim that the longitudinal architecture drives the gain would be refuted, because the gain would then be an artifact of information disparity.
Extended reading notes
Core claim
The central claim is that ETR prediction is a longitudinal regression problem, and that modeling the update sequence directly with a transformer is what improves performance. Concretely, the paper reports that the Longitudinal Tabular Transformer achieves the best Weighted Asymmetric Error and Customer Satisfaction Impact among all tested methods on every storm-size category for all three operating companies, with an average CSI improvement of 19.08% over the baselines. It also introduces the CSI metric itself, defined as $1 - (\alpha \cdot \mathrm{UPR} + \beta \cdot \mathrm{OPR\text{-}8})/(\alpha+\beta)$, with $\alpha=5$ and $\beta=2$ set from customer-survey research, and proposes an asymmetric loss whose piecewise form matches the measured satisfaction curve: under-predictions are punished hardest, small over-predictions up to eight hours are punished least, and large over-predictions are punished again.
Load-bearing premise
The results assume the baseline models were given a fair, information-matched input; the paper says baselines were treated as conventional static tabular regression problems while the LTT saw the full update sequence, and it never clarifies whether baselines saw only the first update or a less informative snapshot.
Editorial extensions
If this is right
- If the reported CSI gain holds, utilities can replace manual or static ETR estimates with predictions that refresh whenever an outage update arrives, reducing the rate at which customers are told restoration will happen sooner than it actually does.
- The asymmetric loss and the eight-hour tolerance give an explicit, tunable link between a regression objective and customer satisfaction; the same piecewise loss could be used for any service-recovery prediction with asymmetric tolerance.
- The feature-attribution and attention results imply that early updates—customer count and priority, then crew-dispatch status and recent restoration rolling averages—carry most of the signal, giving operations teams a concrete list of which field data to keep accurate.
- The architecture's modularity over static inputs, dynamic variables, and variable-length sequences suggests it can be adapted to other longitudinal tabular regression problems without changing the core attention mechanism.
Reading between the lines
- Editorial inference: the same longitudinal framing is a natural fit for water, gas, or telecom outage restoration, where similar update streams exist; the paper does not test those domains.
- Editorial inference: the size of the claimed advantage may depend on the specific tolerance curve—the eight-hour threshold and the $\alpha:\beta$ ratio—so utilities with different customer-survey curves would need to recalibrate before seeing similar gains.
- Editorial inference: a decisive test of the architectural claim would be to give a sequence-capable baseline, such as an LSTM, the same full update history; the paper's comparison treats baselines as static models, so part of the gain could come from access to later updates rather than from self-attention per se.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a Longitudinal Tabular Transformer (LTT) for predicting estimated times of restoration (ETR) from sequences of outage-event updates, using data from three utilities covering about 34,000 storm-related events and 200,000 updates. To reflect customer satisfaction, the authors introduce an asymmetric loss and three associated metrics (UPR, OPR-8, WAE, CSI), and report that LTT improves the CSI metric by an average of 19.08% over eight baselines. They also provide qualitative attention-map and SHAP analyses to interpret temporal dependencies and feature contributions.
Significance. If the reported results can be established under a fair comparison, the work addresses a practically important operational problem and offers a sensible longitudinal treatment of outage updates. The integration of customer-satisfaction considerations into the loss and metrics is useful and could influence industrial practice. However, the current evaluation does not yet support the central claim: the baselines receive an unspecified static input, the reported metrics partly coincide with the training objective, and no statistical significance or variability is reported. These issues must be resolved before the claimed 19.08% CSI improvement can be considered credible.
major comments (3)
- [Baseline Comparisons / Results and Discussion] The manuscript does not state what static input the baseline models receive. The text in 'Baseline Comparisons' says baselines 'process only static, one-dimensional data,' and in 'Results and Discussion' says they are 'treated as a conventional tabular regression problem,' but it never specifies whether the snapshot is the first update, the most recent update, or a fixed time horizon, nor whether baselines are trained on one row per event while LTT is trained on every update prefix. Given that the data include about 200,000 updates for 34,000 events, LTT may be seeing all updates and many more training examples than the baselines. The reported CSI gains could therefore stem from an information and data-volume advantage rather than from the longitudinal architecture. This is the load-bearing comparison for the abstract's headline claim and must be clarified and, if necessary, re-run with information-matched baselines.
- [Evaluation Metrics and Asymmetric Loss (Eqs. 3, 7, 8)] The training loss (Eq. 3), the WAE metric (Eq. 7), and the CSI metric (Eq. 8) share the same piecewise alpha/beta structure. Since LTT is explicitly trained to minimize this exact loss, reporting that LTT has lower WAE and higher CSI than baselines is partly a tautology; the baselines, unless trained with the same loss, are being compared on a metric that the proposed model directly optimizes. The paper should demonstrate improvements on independent metrics that are not part of the training objective, and should report standard deviations or confidence intervals for the five-trial averages in Table 1, along with a valid significance test. The abstract's notation '(p > 0.001)' is not a meaningful significance statement; if a p-value was computed, report the actual value and the test used.
- [Experimental Setup / Data] The data split is described as stratified sampling based on storm magnitude, but the paper does not state whether the split is performed at the event level or the update level. If updates from the same outage event appear in both training and test sets, the longitudinal model could memorize event-specific restoration outcomes, inflating its apparent performance. Please clarify that all updates of a given event are confined to a single split, and describe how the temporal ordering of events is handled to prevent leakage.
minor comments (5)
- [Figure 1 caption] The caption contains a typo: 'Storm-Realted' should be 'Storm-Related'.
- [SHAP Analysis section] In the sentence 'OPCO-3's LLT model predictions,' 'LLT' should be 'LTT' to match the model name used elsewhere.
- [Abstract] The phrase '(p>0.001)' is self-contradictory as written; p>0.001 indicates non-significance, whereas the surrounding context implies a significant difference. If a test was performed, report the exact p-value or a confidence interval.
- [Eq. (10)] The time delta is computed relative to the first update, so the first update always has delta 0. Please clarify whether this is intentional and how it interacts with the positional encoding for the first update.
- [Attention and SHAP interpretation] The attention visualizations and SHAP feature-importance discussion are qualitative. Consider providing quantitative summaries, such as attention concentration metrics or rank correlations across revisions, to support the claimed temporal patterns.
Circularity Check
WAE is the training loss by construction (Eq. 2 = Eq. 6), so the reported WAE advantage and, to a lesser degree, the CSI advantage are partly restatements of the model's own objective rather than independent evidence.
-
self definitional
[Evaluation Metrics, Eq. (6)-(7), compared with Asymmetric Loss Function, Eq. (2)-(3); Results and Discussion]
"This metric is derived from the asymmetric loss function and penalizes under-predictions more heavily than over-predictions... It is defined as: W AE=1/N Σ ℓ(ε_i) where ℓ(ε_i) is the same piecewise loss function used in the asymmetric loss function."
Eq. (6) is identical to Eq. (2): WAE is exactly the training loss L, since both are 1/N Σ ℓ(ε_i) with the same piecewise ℓ(ε_i). The LTT model is trained to minimize this loss (Eq. 2). The paper then reports 'the LTT model's WAE remains the lowest among all tested models' as evidence of robustness. That claim is, up to optimization error, a restatement of the training objective rather than an independent evaluation of the longitudinal architecture. The CSI metric (Eq. 8) uses the same α=5 and β=2 and the same error regions (under-prediction and over-prediction beyond 8 hours) as the loss, so the headline 19.08% CSI improvement is also aligned with the objective; CSI is not literally the loss because it uses rates rather than error magnitudes, giving it some independent content.
full rationale
The paper is not built on a self-citation chain or an imported uniqueness theorem; the only citations used as evidence (e.g., J.D. Power for α and β) are external. The central confound identified here is metric-objective alignment: the WAE evaluation metric is, by the paper's own equations, the same function used to train the model. Reporting WAE as a customer-centric improvement therefore does not by itself validate the transformer architecture; it mainly confirms that the model minimized its own target. The CSI metric is not identical to the loss, since it penalizes the rates of under-prediction and >8-hour over-prediction rather than their magnitudes, so the main headline claim retains some independent content. A separate, non-circularity concern is that baseline models were 'treated as a conventional tabular regression problem' with 'static, one-dimensional data' while LTT received the full update sequence; the paper never specifies which snapshot the baselines saw, which could confound the architecture comparison with an information/volume advantage. That is an evaluation-fairness issue rather than a circularity, so it does not by itself raise the circularity score. Overall, the partial circularity from WAE being the training loss, plus the correlated CSI construction, justifies a moderate score around 5; the paper is not a fully circular derivation because the empirical ranking on held-out storms could still disfavor LTT despite the aligned objective.
Assumptions & free parameters
free parameters (3)
- Alpha (underprediction penalty weight) =
5
- Beta (overprediction-beyond-8h penalty weight) =
2
- LTT hyperparameters (embedding dim, layers, heads, max sequence length) =
128, 6, 16, 20
assumptions (4)
- domain assumption Customer satisfaction depends on ETR prediction error asymmetrically: underprediction is strongly bad, overprediction up to 8 hours is acceptable, and overprediction beyond 8 hours is bad.
- domain assumption Alpha=5 and beta=2 from JD Power research apply across OPCO-1, OPCO-2, and OPCO-3 and across all storm sizes.
- domain assumption A static tabular baseline is a fair and sufficient comparator for the longitudinal model.
- domain assumption The anonymized dataset is representative and free of leakage between training and test storms.
Cite this review
Pith. "Pith review of Predicting Estimated Times of Restoration for Electrical Outages Using Longitudinal Tabular Transformers." pith.science (2026). https://pith.science/paper/52NP3TFE
@misc{pith2026250500225,
author = {Pith},
title = {Pith review of: Predicting Estimated Times of Restoration for Electrical Outages Using Longitudinal Tabular Transformers},
year = {2026},
howpublished = {\url{https://pith.science/paper/52NP3TFE}},
note = {Machine review of arXiv:2505.00225}
}
read the original abstract
As climate variability increases, the ability of utility providers to deliver precise Estimated Times of Restoration (ETR) during natural disasters has become increasingly critical. Accurate and timely ETRs are essential for enabling customer preparedness during extended power outages, where informed decision-making can be crucial, particularly in severe weather conditions. Nonetheless, prevailing utility practices predominantly depend on manual assessments or traditional statistical methods, which often fail to achieve the level of precision required for reliable and actionable predictions. To address these limitations, we propose a Longitudinal Tabular Transformer (LTT) model that leverages historical outage event data along with sequential updates of these events to improve the accuracy of ETR predictions. The model's performance was evaluated over 34,000 storm-related outage events from three major utility companies, collectively serving over 3 million customers over a 2-year period. Results demonstrate that the LTT model improves the Customer Satisfaction Impact (CSI) metric by an average of 19.08% (p > 0.001) compared to existing methods. Additionally, we introduce customer-informed regression metrics that align model evaluation with real-world satisfaction, ensuring the outcomes resonate with customer expectations. Furthermore, we employ interpretability techniques to analyze the temporal significance of incorporating sequential updates in modeling outage events and to identify the contributions of predictive features to a given ETR. This comprehensive approach not only improves predictive accuracy but also enhances transparency, fostering greater trust in the model's capabilities.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
-
[1]
Adibi, M.; and Milanicz, D. 1999. Estimating restoration duration. IEEE Transactions on Power Systems, 14(4): 1493--1498
work page 1999
-
[2]
Chen, T.; and Guestrin, C. 2016. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 785--794
2016
-
[3]
Geno, E. 2023. 2023 Electric Utility Residential Customer Satisfaction Study J . D . Power . https://www.jdpower.com/business/press-releases/2023-electric-utility-residential-customer-satisfaction-study
work page 2023
-
[4]
Gorishniy, Y.; Rubachev, I.; Khrulkov, V.; and Babenko, A. 2021. Revisiting deep learning models for tabular data. Advances in Neural Information Processing Systems, 34: 18932--18943
2021
-
[5]
He, K.; Zhang, X.; Ren, S.; and Sun, J. 2016. Deep Residual Learning for Image Recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770--778
2016
-
[6]
Huang, X.; Khetan, A.; Cvitkovic, M.; and Karnin, Z. 2020. Tabtransformer: Tabular data modeling using contextual embeddings. arXiv preprint arXiv:2012.06678
arXiv 2020
-
[7]
Juan, P. C.; and Joan, C. 2024. Weather-Related Power Outages Rising Climate Central . Technical report, Lawrence Berkeley National Laboratory (United States)
work page 2024
-
[8]
A.; Turner, B.; Levinson, S.; Chen, Y.; Roberts, N.; Prica, M.; and Iyengar, A
Kar, B.; Brewer, J.; Omitaomu, O. A.; Turner, B.; Levinson, S.; Chen, Y.; Roberts, N.; Prica, M.; and Iyengar, A. 2022. RePOWERD: Restoration of Power Outage from Wide-area Severe Weather Disruptions. Technical report, Oak Ridge National Lab.(ORNL), Oak Ridge, TN (United States)
work page 2022
Show all 16 references
-
[9]
Ke, G.; Meng, Q.; Finley, T.; Wang, T.; Chen, W.; Ma, W.; Ye, Q.; and Liu, T.-Y. 2017. Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems, 30
2017
-
[10]
B.; et al
Kinga, D.; Adam, J. B.; et al. 2015. A method for stochastic optimization. In International conference on learning representations (ICLR), volume 5, 6. San Diego, California
2015
-
[11]
M.; and Lee, S.-I
Lundberg, S. M.; and Lee, S.-I. 2017. A Unified Approach to Interpreting Model Predictions. In Guyon, I.; Luxburg, U. V.; Bengio, S.; Wallach, H.; Fergus, R.; Vishwanathan, S.; and Garnett, R., eds., Advances in Neural Information Processing Systems 30, 4765--4774. Curran Asso...
2017
-
[12]
V.; and Gulin, A
Prokhorenkova, L.; Gusev, G.; Vorobev, A.; Dorogush, A. V.; and Gulin, A. 2018. CatBoost: unbiased boosting with categorical features. Advances in neural information processing systems, 31
2018
-
[13]
C.; Wanik, D
Walsh, T. C.; Wanik, D. W.; Anagnostou, E. N.; and Mellor, J. E. 2020. Estimated time to restoration of hurricane sandy in a future climate. Sustainability, 12(16): 6502
2020
-
[14]
Wanik, D.; Anagnostou, E.; Hartman, B.; and Layton, T. 2018. Estimated time of restoration (ETR) guidance for electric distribution networks. Journal of Homeland Security and Emergency Management, 15(1): 20160063
2018
-
[15]
, " * write output.state after.block = add.period write newline
ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...
-
[16]
write newline
" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...
Reviewed August 16, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.