Pith. sign in

REVIEW 4 major objections 5 minor 43 references

CCE: Confidence-Consistency Evaluation for Time Series Anomaly Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper proposes CCE, an anomaly-detection evaluation metric that multiplies prediction confidence by a Bayesian uncertainty-consistency term, and claims it is the only interval-level metric that ranks models perfectly, survives score no

desk verdict A useful TSAD evaluation metric with O(n) complexity and boundedness, but the headline validation is circular and the robustness proof has real gaps. read the letter →

arxiv 2509.01098 v1 pith:BLLCBCT5 submitted 2025-09-01 cs.LG cs.AIstat.ML

Zhijie Zhong , Zhiwen Yu , Yiu-ming Cheung , Kaixiang Yang This is my paper · ORCID
classification cs.LGcs.AIstat.ML
keywords timeseriesanomalydetectionevaluationmetricBayesianuncertaintyestimationBetadistributionrankingbenchmarkLipschitzrobustnessmodelconfidence
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that how an anomaly detector's scores are shaped matters as much as whether they match the labels, and proposes CCE, a metric that scores a detector on both confidence and the consistency of its uncertainty. The metric is built by treating each event's scores as a Beta distribution, reading the distribution's variance as uncertainty, and converting that uncertainty into a penalty term; the authors prove the resulting score is bounded in [0,1], Lipschitz-robust to score perturbations, and computable in O(n) time. They also introduce RankEval, a benchmark that ranks the ranking metrics themselves, using Spearman, Kendall, and mean-rank-deviation against ground-truth model quality. On that benchmark, CCE is the only metric with perfect ranking consistency across every scenario and the only one insensitive to injected noise, which would make it the first fast, threshold-free, uncertainty-aware standard for choosing time series anomaly detection models.

What carries the argument

The key machinery is the Beta-distribution uncertainty estimate. The paper treats each event's normalized anomaly scores as draws from a Beta(alpha, beta), with alpha and beta fixed by the sample mean and second central moment (method of moments), and takes the Beta variance U = alpha*beta / ((alpha+beta)^2 (alpha+beta+1)) as the event's uncertainty. Consistency of an event is exp(-mean U), so high-variance score regions are down-weighted even when their mean score is high; the same U-derived bounds (max variance 1/4, and the constant C1 = 1.5 in the robustness proof) underlie both the boundedness and perturbation-robustness theorems.

What would settle it

Run a Kolmogorov-Smirnov test comparing each event's empirical score distribution to the fitted Beta(alpha, beta) from Eq. (3) on outputs of standard detectors (e.g., LSTMAD, USAD, AnomalyTransformer) across MSL, SMD, PSM, and SWaT; if a substantial fraction of events reject the Beta fit, the consistency term is not measuring what the proofs assume. A direct ranking test also works: construct two detectors with identical expected accuracy but score distributions of different shape (bimodal versus Beta-like); CCE predicts the lower-variance one ranks better, and that ordering is checkable again

Watch

Extended reading notes

Core claim

The paper claims that evaluating a time series anomaly detector should not stop at whether its scores match the labels; the scores themselves carry a reliability signal. CCE models each normal and anomalous event's scores as a Beta distribution (parameters via method of moments), reads the distribution's variance as the event's uncertainty, and derives an event-level consistency score exp(-mean uncertainty). Multiplied by a confidence score (mean score minus a symmetric threshold) and combined across events and the whole series, this yields one number in [0,1] that the authors prove is bounded, Lipschitz-continuous in the scores with constant near 1, and computable in O(n). On the paper's ow

Load-bearing premise

The load-bearing premise is that each event's anomaly scores are well approximated by a Beta distribution, because the consistency penalty is computed from Beta-variance and the Lipschitz proof is derived from Beta-specific derivatives; if real scores are multimodal or heavy-tailed, both the uncertainty's meaning and the robustness bound can break down.

Editorial extensions

If this is right

  • If CCE is right, model selection in time series anomaly detection can use a threshold-free metric whose ranking matches ground-truth model quality without tuning buffers or thresholds, unlike Aff-F1, VUS-ROC, and PATE.
  • CCE's O(n) cost makes interval-aware evaluation feasible on million-point industrial series, where the paper measures VUS-ROC and PATE as orders of magnitude slower.
  • The consistency term introduces a new signal: models that emit confident-looking but erratic scores are penalized even when their thresholded labels look fine, making prediction reliability itself a measurable and trainable property.
  • The RankEval pipeline makes metric-versus-metric comparison reproducible, so future proposed metrics can be ranked against existing ones rather than defended on single datasets.
  • Because CCE is symmetric between normal and anomalous events, it can report which side of a detector's behavior is weaker, guiding targeted model improvement.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The Beta assumption is untested: since the consistency term inherits all its meaning from the variance of a fitted Beta, real score distributions that are multimodal or heavy-tailed could make CCE's uncertainty estimate diverge from genuine predictive reliability; a goodness-of-fit check on standard detectors' outputs would show how far the Lipschitz guarantee actually carries.
  • Because CCE consumes only labels and scores, the same confidence-times-consistency product could transfer to ranking detectors in other domains, such as classification calibration or image anomaly detection, where threshold-free uncertainty-aware ranking is also missing.
  • The confidence and consistency terms are both monotone functions of the same Beta variance, so part of CCE's reported edge may come from amplifying score spread rather than from true calibration; ablating the consistency term against a variance-free surrogate (e.g., discretized-score entropy) would separate the two effects.
  • RankEval's synthetic space could be extended with distribution shift, drift, or seasonal anomalies to test whether CCE's perfect ranking and noise immunity survive when the score-generating process itself changes over time.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The manuscript introduces CCE, a TSAD evaluation metric that combines event-level and global confidence scores based on thresholded means of anomaly scores with a consistency term exp(-U), where U is the variance of a Beta distribution fitted to each event by the method of moments. The authors prove boundedness, Lipschitz robustness, and O(n) complexity for CCE, and introduce RankEval, a benchmark with synthetic ASGM score generators and real-world datasets. They report that CCE achieves perfect ranking consistency (Sp=Kd=1.000, MD=0.000) across all RankEval scenarios and is the only metric with complete noise robustness.

Significance. The idea of an uncertainty-aware, threshold-independent interval-level TSAD metric is timely, and the open-source release plus the attempt to standardize metric evaluation through RankEval are useful contributions. If the theoretical properties and ranking results held, CCE would be a practical and efficient evaluation tool. However, the central ranking claim is undermined by the way RankEval constructs ground-truth rankings from the same statistics CCE uses, and the Lipschitz robustness proof contains a derivative bound that is not generally valid. The contribution is a promising first step, but the main claims need substantial revision before publication.

major comments (4)
  1. [Section IV-B, Eqs. (31)-(36)] The proof of Theorem 6 relies on Eq. (33), which states |∂p/∂s_k|≤2/|E_i| and |∂n/∂s_k|≤2/|E_i|. The first bound is loose but harmless (p=α/(α+β)=bar{s}_{E_i}, so the true derivative is 1/|E_i|). The second is incorrect: from n=s(1-s)/m_2−1 with m_2 the sample second central moment, ∂n/∂s_k=(1−2s)/m_2 − 2s(1−s)(s_k−s)/(|E_i| m_2^2), which is unbounded as m_2→0. Thus the uniform constant C1=1.5 in Eq. (36) and L=0.75/|E_i|+1 in Eq. (41) are not established. The premise that real events have length ≥20 is also contradicted by Table S5 (Creditcard segment lengths 1–5). The Lipschitz claim may hold on a restricted domain, but the proof as written does not support it.
  2. [Section V-C1 and Table I] The RankEval expected ranking R* is defined directly from the ASGM generation parameters q and p. In AccQ(-R), the anomaly-event mean score is an increasing function of q, and in PreQ-NegP-P, the normal-event mean is an increasing function of p. CCE's confidence terms (Eqs. (5)–(8)) are threshold functions of exactly these event-level means, and Cons(E_i) in Eq. (9) is a positive factor in [e^{−0.25},1]. Therefore the ranking induced by CCE is essentially the ranking by score mean, i.e., the same statistic used to construct R*. The perfect Avg. Sp=1.000, Kd=1.000, MD=0.000 in Table I is a built-in consistency check rather than independent evidence that CCE is the best ranking metric. Independent benchmarks with ground truth not monotonically related to the mean score, or real model outputs with externally judged quality, are needed. The ASGM definitions in Sec. V-A2 are also self-contrad
  3. [Section III-C, Algorithm 1, Table III] The confidence definitions use max(·,0) (Eqs. (5), (7)), and Algorithm 1 returns SCCE computed from these definitions, which is nonnegative. However, Table III reports negative CCE scores (ECG LOF −0.01, A.T. −0.21; Power IForest −5.1), which can only arise from the 'relaxed' confidence forms in Eqs. (6) and (8). The paper does not state that experiments use the relaxed variant; this ambiguity affects the interpretation of Table I and all real-data comparisons, and the reported negative scores are inconsistent with the stated bound SCCE∈[0,1]. Please specify exactly which variant was implemented, justify its use, and align the definitions, algorithm, and experiments.
  4. [Section III-B and IV-B] The entire consistency term and the Lipschitz proof depend on the assumption that anomaly scores for each event follow a Beta distribution and that the method-of-moments parameters in Eq. (3) are meaningful. No goodness-of-fit or sensitivity analysis is provided for real TSAD scores, which are often skewed, multimodal, or degenerate (e.g., constant segments). In particular, the derivative bounds in the robustness proof use the Beta variance formula (Eq. (4)) and fail or require regularization when m_2=0. The paper should either prove the property under weaker assumptions on the score distribution, add a robustness check on distributions other than Beta, or state the restricted domain. Additionally, Theorem 6 only analyzes event-level scores; the global scores used in Eq. (13) are not bounded by the same argument, although Eq. (40) applies the event-level Lipschitz constant to the sum.
minor comments (5)
  1. [Algorithm 1] The algorithm lists a 'scale parameter γ' that is never defined or used anywhere in the paper; please remove it or clarify its role.
  2. [Table IV] The header is confusing: the first row reads 'Threshold 0.1 0.3 ...' and the second row reads 'τ 1/1/0 1/1/0 ...'. These should be labeled Sp/Kd/MD for each threshold value.
  3. [Theorem 4, Eqs. (19)-(20)] The notation swaps Conf and Conf*: Eq. (19) uses Conf* for the constrained case and Eq. (20) uses Conf for the relaxed case, opposite to Definitions 3 and 4. Please fix the asterisk convention.
  4. [Section V-D2] The statement that 'eTaPR always considers the model score to be 0' is unclear and should be substantiated with a concrete example or explanation of the metric behavior.
  5. [Section V-A2, AccQ definition] In the AccQ generator, the normal-label case says 'when correctly predicted, the anomaly score is s(1)', but s(1) is the high anomaly score; this is presumably a typo (should be s(2)). Please correct because the synthetic benchmark's semantics depend on this definition.

Circularity Check

1 steps flagged · score 6.0 of 10

RankEval's expected ranking is defined by ASGM parameters q/p, and CCE's confidence is a thresholded mean of the same generated scores, so Table I's perfect scores are built in rather than independent validation.

  1. self definitional [Section V-C1 (Effectiveness Verification), Eqs. (5), (7), (9), Table I]
    "For synthetic tasks, anomaly scores are generated by explicitly given parameters (q, p) in ASGM. Therefore, we can directly obtain the expected performance ranking of each model under ideal conditions based on (q, p) and form the expected ranking R*. ... PreQ-NegP-P only examines false positive rate p, with R* arranged in descending order by p (lower false positive rate is better), ignoring the other dimension."

    R* is not an external ground truth: it is a sort of the ASGM parameters q/p. The generated scores have event-level means that are monotone functions of those same parameters (AccQ anomaly-event mean = 0.025+0.925q; PreQ-NegP normal-event mean = 0.05+0.5p). CCE's confidence terms are thresholded event means (Eqs. 5,7), and the consistency factor (Eq. 9) is bounded in [e^-0.25,1]. Hence, in these synthetic tasks, CCE's ranking is the ranking by score mean, which is R* by construction. Table I's perfect Avg Sp=1.000/Kd=1.000/MD=0.000 therefore shows CCE inherits the generative statistic rather than beating other metrics on an independent benchmark.

full rationale

The load-bearing circularity is in the RankEval validation of the central 'CCE is the best ranking metric' claim. The benchmark's expected ranking is defined by sorting the ASGM generative parameters, and CCE's confidence terms are thresholded versions of the same event-level score means that those parameters control; the consistency term is only a bounded multiplicative factor. Thus the perfect Table I scores are substantially built in. The theoretical Lipschitz and O(n) proofs are not circular, and the real-data visualizations are independent but small and partly qualitative. The Beta-distribution assumption is an unvalidated modeling assumption that affects correctness, but it is not itself a circular step. No load-bearing self-citation issue was found.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The metric's central innovation rests on the Beta distribution assumption, which is an ad hoc modeling choice. The free parameters tau, alpha, and eta are not fitted but are user-specified weights that can affect scores, though the paper argues they do not affect ranking on their synthetic tasks. The synthetic ASGM construction embeds the ground truth in the score distribution parameters, which creates a built-in correlation between CCE's confidence term and the expected ranking. The theoretical robustness proof relies on assumptions about the Beta parameters (n >= 1) that may not hold for the method-of-moments estimates, and the norm mismatch between the L2 statement and the L1 proof is an unverified technical step.

free parameters (5)
  • tau (confidence threshold) = 0.5 (default)
    Used in Eqs. (5)-(8) to define confidence for anomaly and normal events. The paper claims it is dataset-independent but does not provide a principled criterion for choosing it; it is a free hyperparameter. The ablation in Table IV only varies it on the AccQ task, which is a self-consistent scoring scenario.
  • alpha (event-level weight) = 0.5
    Balances anomaly and normal event scores in Eq. (12). Treated as a default with no tuning, but it is an arbitrary weighting choice. Its value can change CCE scores, though the paper claims ranking is stable under tau variations; no analysis of alpha variation is provided.
  • eta (global weight) = 0.5 (inferred)
    Balances global anomaly and normal scores in Eq. (13). The paper does not state an explicit default or sensitivity analysis for eta, making it an implicit free parameter.
  • ASGM accuracy q and precision p = ranges p in [0.1,1.0], q in [0.01,0.3]
    These are not fitted to data but are parameters of the synthetic models used to generate the expected ranking R*. They are free parameters of the experimental design, and the expected ranking is defined by them, so in the synthetic experiments the benchmark is essentially controlled by these values.
  • Gaussian noise sigma = 0.0, 0.05, 0.1
    Used to create ASGM-R models for robustness testing; these are arbitrary noise levels, not derived from real data. The robustness conclusions depend on the chosen noise levels, which are mild.
assumptions (5)
  • ad hoc to paper Anomaly scores for each event follow a Beta distribution with parameters determined by the method of moments.
    This is the key modeling assumption in Section III-B1 (Eq. (2)-(3)). It is not derived from first principles or validated on real anomaly score distributions, yet it underlies the uncertainty term and the robustness proof. If the actual score distribution is not Beta-like, the consistency term is not a meaningful measure of prediction reliability.
  • domain assumption The expected ranking R* in RankEval is the true ranking of model performance, and it is fully determined by the synthetic generation parameters (q, p).
    Used in Section V-C1 to define the ground truth ranking for evaluating metrics. This assumes that accuracy q and precision p are the only factors that determine model quality, ignoring other aspects like calibration or false positive distribution, which could matter in real applications.
  • ad hoc to paper Confidence and consistency are independent enough to be multiplied in event-level and global scores (Eqs. (10)-(14)).
    The product form is a modeling choice, not a derived consequence. The paper provides no justification that confidence (mean score) and consistency (exp of negative variance) are orthogonal or that their product captures model quality better than, say, a sum or a ratio. This is a structural assumption of the metric.
  • domain assumption Normal scores and anomaly scores can be symmetrically evaluated with thresholds symmetric about 0.5.
    Eqs. (5) and (7) assume tau and 1-tau are proper thresholds for anomaly and normal events respectively. This assumes scores are calibrated probabilities, but raw anomaly scores from many detectors are not probabilities and are often not well-calibrated. The paper normalizes scores to [0,1] but does not calibrate them.
  • ad hoc to paper The L2 -> L1 norm bound used in the Lipschitz proof is valid for the perturbation analysis.
    In Eq. (24) the paper bounds |Conf(E_i) - Conf(E~_i)| <= (1/|E_i|) * sum |s_k - s~_k| <= delta, but the assumption stated in Theorem 6 is ||s - s~||_2 <= delta, not an L1 bound. Converting between norms would require an extra factor of sqrt(n), so the stated Lipschitz constant L = 0.75/|E_i| + 1 is not justified by the stated assumptions.
invented entities (1)
  • None
    purpose: No new physical or mathematical entities are invented.
    The paper introduces no new particles, forces, or conserved quantities. It proposes a new metric and a benchmark, which are not entities in the sense of needing independent falsifiable handles.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CCE: Confidence-Consistency Evaluation for Time Series Anomaly Detection." pith.science (2026). https://pith.science/paper/BLLCBCT5

@misc{pith2026250901098,
  author       = {Pith},
  title        = {Pith review of: CCE: Confidence-Consistency Evaluation for Time Series Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BLLCBCT5}},
  note         = {Machine review of arXiv:2509.01098}
}
abstract

Time Series Anomaly Detection metrics serve as crucial tools for model evaluation. However, existing metrics suffer from several limitations: insufficient discriminative power, strong hyperparameter dependency, sensitivity to perturbations, and high computational overhead. This paper introduces Confidence-Consistency Evaluation (CCE), a novel evaluation metric that simultaneously measures prediction confidence and uncertainty consistency. By employing Bayesian estimation to quantify the uncertainty of anomaly scores, we construct both global and event-level confidence and consistency scores for model predictions, resulting in a concise CCE metric. Theoretically and experimentally, we demonstrate that CCE possesses strict boundedness, Lipschitz robustness against score perturbations, and linear time complexity $\mathcal{O}(n)$. Furthermore, we establish RankEval, a benchmark for comparing the ranking capabilities of various metrics. RankEval represents the first standardized and reproducible evaluation pipeline that enables objective comparison of evaluation metrics. Both CCE and RankEval implementations are fully open-source.

Figures

Figures reproduced from arXiv: 2509.01098 by the authors.

Figure 1
Figure 1. Workflow of CCE framework. Note that the confidence and consis [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Latency distribution of different metrics. The violin plot shows the [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. CCE latency analysis under different scenarios: (a) task type impact, (b) time series length impact, and (c) anomaly segment count impact. [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: (a) Performance of different metrics on the AccQ task, (b) Performance [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 6
Figure 6. Figure 6: Visualization of real-world datasets. strating the robustness limitations of other metrics. Addition￾ally, combined with visualization analysis, the effectiveness and robustness of CCE were further demonstrated. The CCE metric and RankEval benchmark provide a comprehen…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 40 canonical work pages

  1. [1]

    Graph anomaly detection in time series: A survey,

    T. K. K. Ho, A. Karami, and N. Armanfard, “Graph anomaly detection in time series: A survey,” IEEE Trans- actions on Pattern Analysis and Machine Intelligence , vol. 47, no. 8, pp. 6990–7009, 2025

    work page 2025

  2. [2]

    A survey on graph neu- ral networks for time series: Forecasting, classification, imputation, and anomaly detection,

    M. Jin, H. Y . Koh, Q. Wen, D. Zambon, C. Alippi, G. I. Webb, I. King, and S. Pan, “A survey on graph neu- ral networks for time series: Forecasting, classification, imputation, and anomaly detection,” IEEE Transactions JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12 on Pattern Analysis and Machine Intelligence , vol. 46, no. 12, pp. 10 466–1...

    work page 2021

  3. [3]

    Self- supervised learning for time series analysis: Taxonomy, progress, and prospects,

    K. Zhang, Q. Wen, C. Zhang, R. Cai, M. Jin, Y . Liu, J. Y . Zhang, Y . Liang, G. Pang, D. Song, and S. Pan, “Self- supervised learning for time series analysis: Taxonomy, progress, and prospects,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 46, no. 10, pp. 6775–6794, 2024

    work page 2024

  4. [4]

    Online learning and se- quential anomaly detection in trajectories,

    R. Laxhammar and G. Falkman, “Online learning and se- quential anomaly detection in trajectories,” IEEE Trans- actions on Pattern Analysis and Machine Intelligence , vol. 36, no. 6, pp. 1158–1173, 2014

    work page 2014

  5. [5]

    Insightful sim- plicity: Dissimilarity in time series anomaly detection,

    Z. Zhong, Z. Yu, J. Chen, and K. Yang, “Insightful sim- plicity: Dissimilarity in time series anomaly detection,” in Proceedings of the ACM Turing Award Celebration Conference - China 2024 , ser. ACM-TURC ’24. New York, NY , USA: Association for Computing Machinery, 2024, p. 242–243

    work page 2024

  6. [6]

    A New Perspective on Time Series Anomaly Detection: Faster Patch-based Broad Learning System

    P. Li, Z. Zhong, T. Zhang, Z. Yu, C. Chen, and K. Yang, “A new perspective on time series anomaly detection: Faster patch-based broad learning system,”arXiv preprint arXiv:2412.05498, 2024

    work page Pith review arXiv 2024

  7. [7]

    Patchad: A lightweight patch-based mlp-mixer for time series anomaly detection,

    Z. Zhong, Z. Yu, Y . Yang, W. Wang, K. Yang, and C. L. P. Chen, “Patchad: A lightweight patch-based mlp-mixer for time series anomaly detection,” IEEE Transactions on Big Data , pp. 1–15, 2025

    work page 2025

  8. [8]

    TFAD: A Decomposition Time Series Anomaly Detection Architecture with Time-Frequency Analysis

    C. Zhang, T. Zhou, Q. Wen, and L. Sun, “TFAD: A Decomposition Time Series Anomaly Detection Archi- tecture with Time-Frequency Analysis,” in Proceedings of the 31st ACM International Conference on Information & Knowledge Management , Oct. 2022, pp. 2497–2507, arXiv:2210.09693 [cs]

    work page Pith review arXiv 2022

Show all 43 references
  1. [9]

    Calibrated one-class classification for unsupervised time series anomaly detection,

    H. Xu, Y . Wang, S. Jian, Q. Liao, Y . Wang, and G. Pang, “Calibrated one-class classification for unsupervised time series anomaly detection,” IEEE Transactions on Knowl- edge and Data Engineering , vol. 36, no. 11, pp. 5723– 5736, 2024

    2024

  2. [10]

    Imdiffusion: Imputed diffusion models for multivariate time series anomaly detection,

    Y . Chen, C. Zhang, M. Ma, Y . Liu, R. Ding, B. Li, S. He, S. Rajmohan, Q. Lin, and D. Zhang, “Imdiffusion: Imputed diffusion models for multivariate time series anomaly detection,” Proc. VLDB Endow., vol. 17, no. 3, p. 359–372, Nov. 2023

    2023

  3. [11]

    Time evidence fusion network: Multi-source view in long-term time series forecasting,

    T. Zhan, Y . He, Y . Deng, Z. Li, W. Du, and Q. Wen, “Time evidence fusion network: Multi-source view in long-term time series forecasting,” IEEE Transactions on Pattern Analysis and Machine Intelligence , pp. 1–15, 2025

    2025

  4. [12]

    Difformer: Multi-resolutional differencing transformer with dynamic ranging for time series analysis,

    B. Li, W. Cui, L. Zhang, C. Zhu, W. Wang, I. W. Tsang, and J. T. Zhou, “Difformer: Multi-resolutional differencing transformer with dynamic ranging for time series analysis,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 45, no. 11, pp. 13 586– 13 598, 2023

    2023

  5. [13]

    When model meets new normals: Test-time adaptation for unsupervised time- series anomaly detection,

    D. Kim, S. Park, and J. Choo, “When model meets new normals: Test-time adaptation for unsupervised time- series anomaly detection,” Proceedings of the AAAI Con- ference on Artificial Intelligence , vol. 38, no. 12, pp. 13 113–13 121, Mar. 2024

    2024

  6. [14]

    Current time series anomaly detection benchmarks are flawed and are creating the illusion of progress,

    R. Wu and E. J. Keogh, “Current time series anomaly detection benchmarks are flawed and are creating the illusion of progress,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 3, pp. 2421–2429, 2023

    2023

  7. [15]

    Simad: A simple dissimilarity- based approach for time-series anomaly detection,

    Z. Zhong, Z. Yu, X. Xi, Y . Xu, W. Cao, Y . Yang, K. Yang, and J. You, “Simad: A simple dissimilarity- based approach for time-series anomaly detection,” IEEE Transactions on Neural Networks and Learning Systems , pp. 1–12, 2025

    2025

  8. [16]

    Navigating the metric maze: a taxonomy of evaluation metrics for anomaly detection in time series,

    S. Sørbø and M. Ruocco, “Navigating the metric maze: a taxonomy of evaluation metrics for anomaly detection in time series,” Data Min. Knowl. Discov., vol. 38, no. 3, p. 1027–1068, Nov. 2023

    2023

  9. [17]

    Precision and recall for time series,

    N. Tatbul, T. J. Lee, S. Zdonik, M. Alam, and J. Gottschlich, “Precision and recall for time series,” in Advances in Neural Information Processing Sys- tems 31: Annual Conference on Neural Information Processing Systems 2018, NeurIPS 2018, December 3- 8, 2018, Montr ´eal, Cana...

    2018

  10. [18]

    Unsupervised anomaly detection via variational auto-encoder for seasonal kpis in web applications,

    H. Xu, W. Chen, N. Zhao, Z. Li, J. Bu, Z. Li, Y . Liu, Y . Zhao, D. Pei, Y . Fenget al., “Unsupervised anomaly detection via variational auto-encoder for seasonal kpis in web applications,” in Proceedings of the 2018 world wide web conference , 2018, pp. 187–196

    2018

  11. [19]

    Timeseriesbench: An industrial-grade benchmark for time series anomaly de- tection models,

    H. Si, J. Li, C. Pei, H. Cui, J. Yang, Y . Sun, S. Zhang, J. Li, H. Zhang, J. Han et al. , “Timeseriesbench: An industrial-grade benchmark for time series anomaly de- tection models,” 2024 IEEE 35th International Sympo- sium on Software Reliability Engineering (ISSRE) , pp. 61...

    2024

  12. [20]

    Local evalu- ation of time series anomaly detection algorithms,

    A. Huet, J. M. Navarro, and D. Rossi, “Local evalu- ation of time series anomaly detection algorithms,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2022, pp. 635– 645

    2022

  13. [21]

    V olume under the sur- face: a new accuracy evaluation measure for time-series anomaly detection,

    J. Paparrizos, P. Boniol, T. Palpanas, R. S. Tsay, A. El- more, and M. J. Franklin, “V olume under the sur- face: a new accuracy evaluation measure for time-series anomaly detection,” Proceedings of the VLDB Endow- ment, vol. 15, no. 11, pp. 2774–2787, 2022

    2022

  14. [22]

    Pate: Proximity-aware time series anomaly evaluation,

    R. Ghorbani, M. J. Reinders, and D. M. Tax, “Pate: Proximity-aware time series anomaly evaluation,” in Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , ser. KDD ’24. New York, NY , USA: Association for Computing Ma- chinery, 2024, p. 872–883

    2024

  15. [23]

    Broad learning autoencoder with graph structure for data clustering,

    Z. Yu, Z. Zhong, K. Yang, W. Cao, and C. L. P. Chen, “Broad learning autoencoder with graph structure for data clustering,” IEEE Transactions on Knowledge and Data Engineering, vol. 36, no. 1, pp. 49–61, 2024

    2024

  16. [24]

    Humanmod: A multi-rag collaborative llm for inclusive urban public healthcare services,

    S. Sun, Z. Zhong, N. Yu, X. Gong, and K. Yang, “Humanmod: A multi-rag collaborative llm for inclusive urban public healthcare services,” Applied Soft Comput- ing, vol. 184, p. 113684, 2025

    2025

  17. [25]

    Towards efficient anomaly detection using memory broad learning system,

    Z. Zhong, K. Yang, Z. Yu, Y . Shi, and C. L. Philip Chen, JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 “Towards efficient anomaly detection using memory broad learning system,” in 2023 9th International Con- ference on Control Science and Systems Engineering (I...

    2021

  18. [26]

    The elephant in the room: Towards a reliable time-series anomaly detection bench- mark,

    Q. Liu and J. Paparrizos, “The elephant in the room: Towards a reliable time-series anomaly detection bench- mark,” in Advances in Neural Information Processing Systems, A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang, Eds., vol. 37. Curran As...

    2024

  19. [27]

    PVLDB, vol

    TSB-AutoAD: Towards Automated Solutions for Time- Series Anomaly Detection, ser. PVLDB, vol. 18. VLDB Foundation, 2025

    2025

  20. [28]

    Bayesian estimation of beta mixture models with variational inference,

    Z. Ma and A. Leijon, “Bayesian estimation of beta mixture models with variational inference,” IEEE Trans- actions on Pattern Analysis and Machine Intelligence , vol. 33, no. 11, pp. 2160–2173, 2011

    2011

  21. [29]

    Bayesian outlier detection in non-gaussian autoregressive time series,

    M. E. Silva, I. Pereira, and B. McCabe, “Bayesian outlier detection in non-gaussian autoregressive time series,” Journal of Time Series Analysis , vol. 40, no. 5, pp. 631– 648, 2019

    2019

  22. [30]

    Time Series Anomaly Detection Using Transformer-Based GAN With Two-Step Masking,

    A.-H. Shin, S. T. Kim, and G.-M. Park, “Time Series Anomaly Detection Using Transformer-Based GAN With Two-Step Masking,” IEEE Access , vol. 11, pp. 74 035– 74 047, 2023

    2023

  23. [31]

    Landmark block-embedded aggregation autoencoder for anomaly detection,

    Y . Liu, Y . Tian, Y . Mi, H. Liu, J. Wang, and W. Pedrycz, “Landmark block-embedded aggregation autoencoder for anomaly detection,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 55, no. 2, pp. 1004–1019, 2025

    2025

  24. [32]

    The ucr time series archive,

    H. A. Dau, A. Bagnall, K. Kamgar, C.-C. M. Yeh, Y . Zhu, S. Gharghabi, C. A. Ratanamahatana, and E. Keogh, “The ucr time series archive,” IEEE/CAA Journal of Automatica Sinica, vol. 6, no. 6, pp. 1293–1305, 2019

    2019

  25. [33]

    Lof: identifying density-based local outliers,

    M. M. Breunig, H.-P. Kriegel, R. T. Ng, and J. Sander, “Lof: identifying density-based local outliers,” in Pro- ceedings of the 2000 ACM SIGMOD international con- ference on Management of data , 2000, pp. 93–104

    2000

  26. [34]

    Isolation forest,

    F. T. Liu, K. M. Ting, and Z.-H. Zhou, “Isolation forest,” in 2008 eighth ieee international conference on data mining. IEEE, 2008, pp. 413–422

    2008

  27. [35]

    USAD: UnSupervised Anomaly Detection on Multivariate Time Series,

    J. Audibert, P. Michiardi, F. Guyard, S. Marti, and M. A. Zuluaga, “USAD: UnSupervised Anomaly Detection on Multivariate Time Series,” in Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining . Virtual Event CA USA: ACM, 2020, pp. 3395–3404

    2020

  28. [36]

    Anomaly trans- former: Time series anomaly detection with association discrepancy,

    J. Xu, H. Wu, J. Wang, and M. Long, “Anomaly trans- former: Time series anomaly detection with association discrepancy,” in International Conference on Learning Representations, 2021

    2021

  29. [37]

    Timesnet: Temporal 2d-variation modeling for general time series analysis,

    H. Wu, T. Hu, Y . Liu, H. Zhou, J. Wang, and M. Long, “Timesnet: Temporal 2d-variation modeling for general time series analysis,” in The eleventh international con- ference on learning representations , 2022

    2022

  30. [38]

    The proof and measurement of association between two things,

    C. Spearman, “The proof and measurement of association between two things,” The American journal of psychol- ogy, vol. 100, no. 3/4, pp. 441–471, 1987

    1987

  31. [39]

    A new measure of rank correlation,

    M. G. Kendall, “A new measure of rank correlation,” Biometrika, vol. 30, no. 1-2, pp. 81–93, 1938

    1938

  32. [40]

    Ranking of classi- fication algorithms in terms of mean–standard deviation using a-topsis,

    A. G. Pacheco and R. A. Krohling, “Ranking of classi- fication algorithms in terms of mean–standard deviation using a-topsis,” Annals of Data Science, vol. 5, no. 1, pp. 93–110, 2018

    2018

  33. [41]

    Practical ap- proach to asynchronous multivariate time series anomaly detection and localization,

    A. Abdulaal, Z. Liu, and T. Lancewicki, “Practical ap- proach to asynchronous multivariate time series anomaly detection and localization,” in Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 2021, pp. 2485–2494

    2021

  34. [42]

    Detecting spacecraft anomalies us- ing lstms and nonparametric dynamic thresholding,

    K. Hundman, V . Constantinou, C. Laporte, I. Colwell, and T. Soderstrom, “Detecting spacecraft anomalies us- ing lstms and nonparametric dynamic thresholding,” in Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, 2018, pp. 387–395

    2018

  35. [43]

    A dataset to support research in the design of secure water treatment systems,

    J. Goh, S. Adepu, K. N. Junejo, and A. Mathur, “A dataset to support research in the design of secure water treatment systems,” in Critical Information Infrastruc- tures Security, G. Havarneanu, R. Setola, H. Nassopou- los, and S. Wolthusen, Eds. Cham: Springer Interna- tional...

    2017

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.