Pith. sign in

REVIEW 5 major objections 5 minor 51 references

Know Unreported Roadway Incidents in Real-time: Early Traffic Anomaly Detection

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A deep learning system that learns to flag roadway anomalies before they are reported.

desk verdict A genuinely useful labeling pipeline for traffic anomaly detection, but the evaluation is circular: labels, FAR, and significant-incident definitions all derive from the same slowdown-speed input, so the headline early-detection claim is unverified. read the letter →

arxiv 2412.10892 v2 pith:E6REIXPW submitted 2024-12-14 cs.LG cs.AI

Haocheng Duan , Hao Wu , Sean Qian This is my paper · ORCID
classification cs.LGcs.AI
keywords earlytrafficanomalydetectionautomaticincidentslowdownspeedlabeldenoisingaheadlabelingprobevehiclesequence-to-sequencepredictionmanagement
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

The paper's thesis is that road incidents can be known before anyone reports them. The authors build a per-segment deep learning system that ingests only ubiquitous low-cost data—probe-vehicle speeds, weather, and historical incident feeds—and predicts whether a segment will be anomalous over the next half hour. They train it not on raw incident reports but on automatically denoised anomaly labels: incident reports are kept only when they coincide with a spatial slowdown signal, and prolonged slowdown events with no report at all are added as anomalies. On ten highway segments in Maryland and Pennsylvania, the model alerts 5 to 42 minutes before Waze reports, with negative mean time to detection on every segment. If this holds for other networks, traffic operators could act during the minutes that currently elapse between an incident's onset and its first report.

What carries the argument

The load-bearing object is the slowdown speed of a segment, defined as the positive difference between the average speed of upstream segments and the segment's own speed. It is used as a prior-knowledge filter in a two-step labeling algorithm: incident reports are kept as anomaly labels only if a slowdown-speed threshold is exceeded somewhere in their time window, and prolonged slowdown events without any report are added as unreported anomalies. Ahead-labeling then extends each anomaly window 15 minutes backward so early-stage features enter training. The prediction model is a per-segment sequence-to-sequence Transformer/GraphTrans that maps the past hour of sub-graph traffic, weather, and time features onto anomaly status for the next six 5-minute steps, trained with weighted binary cross-entropy, teacher forcing, sliding windows applied after chronological splits, and a separately tuned alert threshold.

What would settle it

On a short segment where incidents typically start at the segment's beginning, compute the overlap between top slowdown-speed timestamps and independently verified incident times; if the overlap falls far below the reported 60 percent and the model's mean time to detection turns positive, the early-detection claim is refuted. A lead over camera-confirmed onset times, rather than report times, would be the cleaner confirmation that the alerts are genuinely predictive.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that early anomaly detection is learnable from probe speed and weather data without relying on incident reports as ground truth. The central empirical claim is that the proposed framework, trained with prior-knowledge-denoised labels and multi-step prediction, detects or predicts incidents before they appear in any report: across all ten test segments the mean time to detection relative to report time is negative, and alerts arrive 5–42 minutes earlier than Waze reports. The paper further claims that the model detects significant incidents at nearly perfect rate while keeping false alarm rates low, and that the denoised labels make training converge where raw incident-report labels do not. The authors interpret negative MTTD as evidence that the model has learned early-stage spatial signatures—upstream and downstream speeds—that precede both the report and the visible speed drop on the target segment.

Load-bearing premise

The whole pipeline assumes that a large slowdown speed—the gap between upstream traffic speed and the segment's own speed—reliably marks real, significant anomalies, and the authors note this holds only for relatively long segments where incidents do not start at the segment's beginning; where that correlation is weak, the generated labels and hence the model are built on a biased signal.

Editorial extensions

If this is right

  • Incident alerts can be issued before any human or crowdsourced report exists, giving operators 5–42 minutes of lead time on the tested segments.
  • Training on denoised anomaly labels rather than raw incident reports makes deep models converge and outperform statistical, SVM, GAN, and autoencoder baselines in F1 and false alarm rate.
  • The method transfers across two different road networks using only data sources that cover most major roads, so it can be deployed without new detectors or manual report filtering.
  • Ahead-labeling shortens mean time to detection in 8 of 10 test cases, at the cost of a small increase in false alarms.
  • A shared-decoder multi-step formulation lets a single model issue alerts for a rolling half-hour horizon, a form of prediction rather than only now-casting.

Reading between the lines

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

  • If the slowdown-speed prior is as transferable as the paper suggests, the same label-denoising recipe could be applied to other anomaly domains with a cheap spatial proxy—for example, water or power network disruptions detected through pressure or voltage differentials.
  • The negative MTTD claims are measured against report times, not true incident onset; verifying against independently camera-confirmed onset times would determine how much of the lead time is real anticipation versus reporting latency.
  • The per-segment model design trades global transferability for simplicity; a single network-wide model that shares the slowdown-speed prior might extend the method to segments with very few incidents, which the paper identifies as a gap.
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

5 major / 5 minor

Summary. This paper proposes a framework for early detection/prediction of traffic anomalies using widely available probe-vehicle speed data, incident reports, and weather. The authors generate 'anomaly labels' from a slowdown-speed threshold (Algorithm 1) and extend them backward in time (Algorithm 2), then train per-segment sequence-to-sequence models (Transformer/GraphTrans) to predict these labels up to 30 minutes ahead. On ten segments in Maryland and Pennsylvania, they report negative mean time to detection (MTTD) relative to Waze reports, significant-incident detection rates DR(S) near 1, and low false alarm rates (FAR), concluding that their model alerts 5–42 minutes earlier than Waze reports.

Significance. If the reported results were obtained against independent ground truth, the framework would be practically valuable: it uses low-cost, ubiquitous data, is fully automated, and explicitly addresses delayed and missing incident reports. The paper also takes care to avoid train-test contamination in sliding-window construction and includes a sensitivity analysis of the labeling threshold. However, the evaluation is circular: the ground-truth labels for both training and testing are derived from the same slowdown-speed feature that is a model input, and the false-alarm definition counts any alarm matching those generated labels as a true positive. Consequently, the headline early-detection claim is not supported by the presented evidence.

major comments (5)
  1. [§4.2, Algorithm 1; §4.5; Eq. (13)] The anomaly labels used for both training and evaluation are generated from the slowdown speed SD_i(t) defined in Eq. (4), which is also a direct input feature of the model (Section 4.1/4.5). Eq. (13) counts an alarm as false only if it overlaps neither an incident report nor a generated anomaly label. A trivial detector that raises an alarm whenever SD_i(t) exceeds the Algorithm 1 percentile threshold would, by construction, match most generated labels and yield high DR(S), low FAR, and negative MTTD relative to delayed Waze reports, without detecting any real unreported incident. The headline results in Table 4 are therefore not evidence of early detection of actual anomalies. An independent ground truth (e.g., verified incident logs, camera/operator records, or a public benchmark such as the cited FT-AED dataset) is needed to validate both the labels and the detection timeliness.
  2. [§4.3, Algorithm 2; §5.3; §6] Ahead labeling extends every anomaly label backward by θ_ahead = 3 steps (15 minutes) before training. Because MTTD is measured against the start of Waze reports (Eq. 12) and the evaluation uses the same ahead-extended labels to define significant incidents (Eqs. 14–15), the model is trained to fire exactly in the extended pre-report window. The conclusion that the model 'alerts 5–42 min earlier than Waze reports' therefore follows by construction from the label definition; it does not demonstrate that the model detects an anomaly before it actually occurs or before it is observable in the input features. Timeliness should be re-evaluated against the actual onset of the anomaly (e.g., verified incident time) rather than against the generated labels.
  3. [§6 (Conclusion)] The authors concede that they 'mainly chose edges with a high correlation between the occurrence of incidents and high slowdown speeds,' and that this correlation requires relatively long segments with incidents not at the segment start. Since the labeling algorithm is based on that same correlation, the selected ten segments are exactly those where the labeling assumption is most favorable. The stated generalizability to 'any size of road segments' (Section 1) is therefore unsupported; results on segments lacking the high correlation, or an alternative labeling mechanism not derived from the model's input features, are needed.
  4. [§5.3, Table 4] The learned baselines (SVM, GAN, TSSAE) are trained on raw incident-report labels, while Ours is trained on the generated anomaly labels. The paper itself observes (Section 5.3) that models trained on incident reports often fail to converge. Thus the comparison largely reflects label quality rather than model architecture. A baseline using the same generated labels and the same input features—e.g., a simple threshold on SD_i(t) or a standard classifier trained on the same labels—should be included to isolate the contribution of the proposed deep learning framework.
  5. [Eq. (13)] The false alarm rate is not a conventional false alarm rate. The numerator counts alarms that do not overlap any incident report or generated anomaly label, and the denominator is the total number of alarms. A model that never triggers has FAR = 0, which is why several baselines in Table 4 with DR = 0 also show FAR = 0. The paper should report a per-time-step false positive rate (e.g., the fraction of non-anomaly time steps with alarms) or precision against independent ground truth.
minor comments (5)
  1. [§4.3, Algorithm 2] The pseudocode's first loop uses 'AANi(p,:q) = 1' for q=1,...,θ_ahead−1, which appears to set a slice from the start rather than the intended q−θ_ahead:q window; the second loop also sets AANi(p,q−θ_ahead:q)=1. Please clarify the exact semantics and loop bounds.
  2. [§4.2 vs §4.3] The text in §4.2 says the prediction interval is 5 minutes, while §4.3 says the prediction interval is 15 minutes; Algorithm 2 uses θ_ahead=3. Please make the temporal units consistent throughout.
  3. [§6 vs Table 4] The conclusion states alerts are 5–42 minutes earlier than Waze reports, but Table 4 shows MTTD values of -4 to -42 for Ours(a), i.e., 4–42 minutes; the range should be reconciled with the table.
  4. [Eq. (1) vs Eq. (12)] Equation (12) uses N_detected_incident while Eq. (1) used N_detected; standardize the notation for the number of detected incidents.
  5. [Table 4] Table 4 contains unusual symbols and crossed-out entries whose meaning is only partially explained; consider a more readable layout with a legend for the strikethrough criteria.

Circularity Check

3 steps flagged · score 8.0 of 10

Headline metrics are scored against labels generated from the same slowdown-speed input the model receives; a threshold baseline on SD could reproduce the near-zero FAR and negative MTTD, so the early-detection claim is unverified.

  1. self definitional [Section 3.2 Eq. (4); Section 4.2 Algorithm 1 (Step 0)]
    "Given that traffic flow data often requires additional detectors and is not widely available, we use the slowdown speed described in Equation (4) as the prior knowledge to infer whether an incident report corresponds to a traffic anomaly. ... obtain the abnormal slowdown speed threshold: θiSD = Percentile(vec(SDi), n); compute the abnormal slowdown speed binary matrix: ASDi(p,q) = I(SDi(p,q) ≥ θiSD)"

    The anomaly labels used as training targets are built from thresholding SD_i(t), which is itself one of the engineered input features (Section 3.2 defines SD; Sections 4.1 and 4.5 feed sub-graph traffic status into the encoder). Thus the target variable is a deterministic function of a model input feature. The paper's disclaimer that 'the model is not trained to detect abnormal slowdown speeds' does not break the loop: ASD events that are prolonged or overlap incident reports are exactly what enters the label, so a threshold classifier on SD reproduces the labels by construction. The Conclusion further admits that segments were chosen for high SD-incident correlation, making such a threshold baseline particularly favorable.

  2. self definitional [Section 5.3.3, Eqs. (13)-(15)]
    "The numerator of FAR has been changed from the number of alarms not coinciding with any incident reports to the number of alarms not coinciding with any incident reports or anomalies. The anomalies here are those generated using the anomaly labels described in Section 4.2."

    FAR, DR(S), and MTTD(S) are evaluated against the same generated labels that were used for training. Because those labels are threshold functions of the SD input, any alarm that coincides with an SD-threshold event is automatically counted as a true positive rather than a false alarm. Consequently, near-zero FAR and DR(S)=1 for 'significant' incidents are largely assured by the evaluation definition, not by independent evidence that unreported incidents were detected. The final comparison to Waze report times is the only external anchor, but true-positive significance is still defined by the internal labels.

1 more flagged steps
  1. self definitional [Section 4.3 Algorithm 2; Section 5.3.3 Table 4]
    "Therefore, we label anomalies starting from a few time steps prior to the reported incident/anomaly time to supplement the lack of early characteristics in the anomaly samples. For example, an anomaly from 7:30 a.m. to 8:30 a.m. can be ahead-labeled by 15 minutes to become from 7:15 a.m. to 8:30 a.m."

    The 15-minute backward extension of every anomaly label is inserted into the training target before the model is trained and before the test metrics are computed. Since FAR counts alarms overlapping these ahead-labeled anomalies as true, and MTTD is measured from the Waze report start, the negative MTTD values and the claimed 5-42 minute advance are at least partly baked into the label definition rather than established against an independent ground truth. No external incident logs or public benchmark are used to verify that the early alarms correspond to real unreported incidents.

full rationale

The circularity here is not a self-citation chain; it is definitional. Algorithm 1 constructs the anomaly ground truth from a percentile threshold on SD_i(t) (Eq. 4), and SD is also an engineered input feature fed into the sub-graph encoder. Algorithm 2 then extends every positive label 15 minutes backward, injecting the 'early' advantage into the target. The evaluation in Section 5.3.3 then scores FAR, DR(S), and MTTD(S) against exactly these generated labels, so the reported near-zero FAR and DR(S)=1 are largely consequences of the evaluation definition. A model that simply raised alerts when current or recent SD exceeded the Algorithm 1 threshold would match the labels and could reproduce the headline results on segments deliberately chosen for high SD-incident correlation, as the Conclusion concedes. The comparison with Waze report times provides only a weak external anchor, because the definition of a true positive is still determined by the internally generated labels. Without independent incident logs or a public benchmark such as the cited FT-AED dataset, the central early-detection claim cannot be distinguished from the model's own label generator. Score 8, not 10, because the model does use additional spatiotemporal features and Waze report times are external, but the central evaluation loop reduces by definition to the slowdown-speed threshold.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the slowdown-speed prior for labeling, the completeness of probe speed data, and the learnability of the generated labels. No code or data are released, and the main performance metrics are defined partly from the same labels the model is trained on.

free parameters (7)
  • slowdown speed percentile n = adjusted per segment via alpha in Algorithm 1; specific values not reported
    Determines theta_i^SD, the threshold that defines 'abnormal slowdown speed'. Tuned to satisfy removal/addition percentage constraints theta1, theta2.
  • removal percentage threshold theta1 = default 0.6
    Controls how many incident reports may be filtered; adjusted slightly per data per Section 4.2.
  • addition percentage threshold theta2 = default 1.0
    Controls how many additional anomalies may be added; adjusted slightly per data.
  • minimum anomaly duration theta_t = default 3 (15 min)
    A slowdown must persist this many steps to be labeled an anomaly.
  • ahead-labeling duration theta_ahead = default 3 (15 min)
    Extends anomaly labels backward in time; paper notes it must be tuned per segment.
  • anomaly weight w_ano in WBCE loss = default ratio of negative to positive samples; further adjusted
    Controls precision/recall trade-off.
  • alert threshold tau = tuned per segment on tuning set, range ~0.4-0.7
    Converts continuous output to binary alarm; chosen by maximizing F1 on tuning set.
assumptions (4)
  • domain assumption A significant traffic incident produces a significant slowdown speed on the target or upstream links (shock-wave theory).
    Used in Algorithm 1 to define anomaly labels; cited to Wirasinghe (1978), but the paper acknowledges it does not hold for all segments.
  • domain assumption Probe-vehicle speed data (INRIX) are complete and accurate enough after imputation to reflect traffic states.
    All features and labels derive from these speeds; imputation uses free-flow speed for missing intervals, which may bias labels.
  • domain assumption Recurrent speed baselines (same weekday, past 3 weeks, excluding incident reports) are stable references for defining anomalies.
    Used in SRS (Equation 6) as an input feature and in case studies to illustrate anomalies; stability is assumed but not quantified.
  • domain assumption Multi-step sequence-to-sequence models can learn early-stage incident features from the generated labels.
    The entire learning framework depends on this assumption; no formal proof or ablative evidence beyond performance tables.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Know Unreported Roadway Incidents in Real-time: Early Traffic Anomaly Detection." pith.science (2026). https://pith.science/paper/E6REIXPW

@misc{pith2026241210892,
  author       = {Pith},
  title        = {Pith review of: Know Unreported Roadway Incidents in Real-time: Early Traffic Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E6REIXPW}},
  note         = {Machine review of arXiv:2412.10892}
}
read the original abstract

This research aims to know traffic anomalies as early as possible. A traffic anomaly refers to a generic incident on the road that influences traffic flow and calls for urgent traffic management measures. `Knowing'' the occurrence of a traffic anomaly is twofold: the ability to detect this anomaly before it is reported anywhere, or it may be such that an anomaly can be predicted before it actually occurs on the road (e.g., non-recurrent traffic breakdown). In either way, the objective is to inform traffic operators of unreported incidents in real time and as early as possible. The key is to stay ahead of the curve. Time is of the essence. Conventional automatic incident detection (AID) methods often struggle with early detection due to their limited consideration of spatial effects and early-stage characteristics. Therefore, we propose a deep learning framework utilizing prior domain knowledge and model-designing strategies. This allows the model to detect a broader range of anomalies, not only incidents that significantly influence traffic flow but also early characteristics of incidents along with historically unreported anomalies. We specially design the model to target the early-stage detection/prediction of an incident. Additionally, unlike most conventional AID studies, our method is highly scalable and generalizable, as it is fully automated with no manual selection of historical reports required, relies solely on widely available low-cost data, and requires no additional detectors. The experimental results across numerous road segments on different maps demonstrate that our model leads to more effective and early anomaly detection.

Figures

Figures reproduced from arXiv: 2412.10892 by the authors.

Figure 1
Figure 1. Incident Management Timeline (Federal Highway Administration (FHWA), 2020a) unreported incidents. This is because they are typically tuned/trained and evaluated based solely on incident reports, thereby inheriting the same delays and missing issues. These issues will be further discussed in detail later in this section. This paper aims to overcome the bottlenecks caused by incident reports in conventional AID method… view at source ↗
Figure 2
Figure 2. Time-varying speeds on a road segment: Waze incident reports were significantly delayed [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Time-varying speeds on a road segment: Incident reports were missing [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: TMC Network and Incident Locations The INRIX speed data used in this study includes all-vehicle speeds at both 1-minute and 5-minute granularities, along with personal vehicle and truck speeds at a 1-minute granularity. All data with 1-minute granularity are imputed to…
Figure 5
Figure 5. Figure 5: Incident Report Count Per Segment incident reports are referenced for anomaly status labeling. To build an effective model, the following issues must be considered. • Rarity and randomness issues of the incident reports: Regarding rarity, it leads to a severe imbalance…
Figure 6
Figure 6. Figure 6: Framework Overview multi-lane highway during off-peak hours might not affect speed or flow, thus resulting in an undetectable incident. This may not be of interest to traffic operators if there is no impact on mobility or safety for the time being. Consequently, using …
Figure 7
Figure 7. Figure 7: Sliding Window and Train-Test Contamination [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Recall, Precision, and F1 score vary significantly across different segments, making them difficult to directly apply in real-world traffic management. To better meet practical needs, we introduce a threshold to determine when to trigger an alert. In other words, we ne…
Figure 9
Figure 9. Figure 9: No Report, Anomaly Labeled 17 [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Reported, Anomaly Labeled 18 [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: Recall, Precision, F1 score, Accuracy Plot (Without Ahead Labelling) [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]
Figure 12
Figure 12. Figure 12: Recall, Precision, F1 score, Accuracy Plot (With Ahead Labelling) [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]
Figure 13
Figure 13. Figure 13: Speed, Alert, and Incident Report on 2023-01-11 [PITH_FULL_IMAGE:figures/full_fig_p025_13.png]
Figure 14
Figure 14. Figure 14: Upstream and Downstream speed on 2023-01-11 [PITH_FULL_IMAGE:figures/full_fig_p025_14.png]
Figure 15
Figure 15. Figure 15: Speed, Alert, and Incident Report on 2023-01-25 [PITH_FULL_IMAGE:figures/full_fig_p026_15.png]
Figure 16
Figure 16. Figure 16: The ’False Positive’ Cases 27 [PITH_FULL_IMAGE:figures/full_fig_p027_16.png]
Figure 17
Figure 17. Figure 17: The ’False Negative’ Cases 28 [PITH_FULL_IMAGE:figures/full_fig_p028_17.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 47 canonical work pages

  1. [1]

    Abdulhai, B., & Ritchie, S. G. (1999). Enhancing the universality and transferability of freeway incident detection using a bayesian-based neural network. Transportation Research Part C: Emerging Technologies, 7 (5), 261–280

    work page 1999

  2. [2]

    Ahmed, F., & Hawas, Y. (2015). An integrated real-time traffic signal system for transit signal priority, incident detection and congestion management. Transportation Research Part C: Emerging Technologies, 60 , 52–76

    work page 2015

  3. [3]

    A., & Cook, A

    Ahmed, S. A., & Cook, A. R. (1982). Application of time-series analysis techniques to freeway incident detection. Transportation Research Record, 841 (3), 19–21

    work page 1982

  4. [4]

    X., & Ushiki, T

    Asakura, Y., Kusakabe, T., Nguyen, L. X., & Ushiki, T. (2017). Incident detection methods using probe vehicles with on-board gps equipment. Transportation research part C: emerging technologies, 81 , 330–341

    work page 2017

  5. [5]

    Bahdanau, D., Cho, K., & Bengio, Y. (2014). Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473

    arXiv 2014

  6. [6]

    L., & Mountain, C

    Balke, K., Dudek, C. L., & Mountain, C. E. (1996). Using probe-measured travel times to detect major freeway incidents in houston, texas. Transportation Research Record, 1554 (1), 213–220

    work page 1996

  7. [7]

    Chakraborty, P., Hegde, C., & Sharma, A. (2019). Data-driven parallelizable traffic incident detection using spatio-temporally denoised robust thresholds. Transportation research part C: emerging technologies, 105 , 81–99

    work page 2019

  8. [8]

    Chakraborty, P., Sharma, A., & Hegde, C. (2018). Freeway traffic incident detection from cameras: A semi-supervised learning approach. In 2018 21st international conference on intelligent transportation systems (itsc) (pp. 1840–1845)

    work page 2018

Show all 51 references
  1. [9]

    L., & Ritchie, S

    Cheu, R. L., & Ritchie, S. G. (1995). Automated detection of lane-blocking freeway incidents using artificial neural networks. Transportation Research Part C: Emerging Technologies, 3 (6), 371–388

    1995

  2. [10]

    L., & Tay, G

    Cheu, R. L., & Tay, G. C. W. (2004). Sampling strategies for probe-vehicle-based freeway incident detection algorithms. Transportation research record, 1867 (1), 80–88

    2004

  3. [11]

    Coursey, A., Ji, J., Quinones Grueiro, M., Barbour, W., Zhang, Y., Derr, T., . . . Work, D. (2024). Ft-aed: Benchmark dataset for early freeway traffic anomalous event detection. Advances in Neural Information Processing Systems , 37 , 15526–15549

    2024

  4. [12]

    Dia, H., & Rose, G. (1997). Development and evaluation of neural network freeway incident detection models using field data. Transportation Research Part C: Emerging Technologies, 5 (5), 313–331

    1997

  5. [13]

    L., Messer, C

    Dudek, C. L., Messer, C. J., & Nuckles, N. B. (1974). Incident detection on urban freeways. Transportation research record, 495 , 12–24

    1974

  6. [14]

    ElSahly, O., & Abdelfatah, A. (2022). A systematic review of traffic incident detection algorithms. Sustainability, 14 (22), 14859. Federal Highway Administration (FHW A). (2020a). Process for establishing, implementing, and institutionalizing a traffic incident management per...

    2022

  7. [15]

    Bengio, Y

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., . . . Bengio, Y. (2014). Generative adversarial nets. Advances in neural information processing systems , 27

    2014

  8. [16]

    Grigorev, A., Mih˘ ait ¸˘ a, A.-S., Saleh, K., & Chen, F. (2024). Automatic accident detection, segmentation and duration prediction using machine learning. IEEE Transactions on Intelligent Transportation Systems, 25 (2), 1547-1568. DOI: 10.1109/TITS.2023.3323636

    2024

  9. [17]

    S., & Chen, F

    Gu, Y., Qian, Z. S., & Chen, F. (2016). From twitter to detector: Real-time traffic incident detection using social media data. Transportation research part C: emerging technologies, 67 , 321–342

    2016

  10. [18]

    C., Li, X., & Sun, W

    Han, X., Grubenmann, T., Cheng, R., Wong, S. C., Li, X., & Sun, W. (2020). Traffic incident detection: A trajectory-based approach. In 2020 ieee 36th international conference on data engineering (icde) (pp. 1866–1869)

    2020

  11. [19]

    H., & Friedman, J

    Hastie, T., Tibshirani, R., Friedman, J. H., & Friedman, J. H. (2009). The elements of statistical learning: data mining, inference, and prediction (Vol. 2). Springer

    2009

  12. [20]

    He, D., Kim, J., Shi, H., & Ruan, B. (2023). Autonomous anomaly detection on traffic flow time series 33 with reinforcement learning. Transportation Research Part C: Emerging Technologies, 150 , 104089

    2023

  13. [21]

    L., & Srinivasan, D

    Jin, X., Cheu, R. L., & Srinivasan, D. (2002). Development and adaptation of constructive probabilistic neural network in freeway incident detection. Transportation Research Part C: Emerging Technologies, 10 (2), 121–147

    2002

  14. [22]

    Karim, A., & Adeli, H. (2002). Incident detection algorithm using wavelet energy representation of traffic patterns. Journal of Transportation Engineering , 128 (3), 232–242

    2002

  15. [23]

    Liu, T.-Y

    Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., . . . Liu, T.-Y. (2017). Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems , 30

    2017

  16. [24]

    Ke, Z., Duan, H., & Qian, S. (2024). Interpretable mixture of experts for time series prediction under recurrent and non-recurrent conditions. arXiv preprint arXiv:2409.03282

    2024 arXiv

  17. [25]

    Ke, Z., Zou, Q., Liu, J., & Qian, S. (2024). Real-time system optimal traffic routing under uncertainties–can physics models boost reinforcement learning? arXiv preprint arXiv:2407.07364

    2024 arXiv

  18. [26]

    Ki, Y.-K., & Lee, D.-Y. (2007). A traffic accident recording and reporting model at intersections. IEEE Transactions on Intelligent Transportation Systems , 8 (2), 188–194

    2007

  19. [27]

    Li, L., Lin, Y., Du, B., Yang, F., & Ran, B. (2022). Real-time traffic incident detection based on a hybrid deep learning model. Transportmetrica A: transport science, 18 (1), 78–98

    2022

  20. [28]

    Lin, Y., Li, L., Jing, H., Ran, B., & Sun, D. (2020). Automated traffic incident detection with a smaller dataset based on generative adversarial networks. Accident Analysis & Prevention , 144 , 105628

    2020

  21. [29]

    Pan, D. (2022). Development of incident detection algorithms compatible with low temporal-resolution traffic detectors and limited incident data (Unpublished doctoral dissertation). The George Washington University

    2022

  22. [30]

    Parkany, E., & Xie, C. (2005). A complete review of incident detection algorithms & their deployment: What works and what doesn ’t (Technical Report No. NETCR37). New England Transportation Consortium. Retrieved from https://onlinepubs.trb.org/onlinepubs/trispdfs/00988875.pdf

    2005

  23. [31]

    B., Movahedi, A., Taghipour, H., Derrible, S., & Mohammadian, A

    Parsa, A. B., Movahedi, A., Taghipour, H., Derrible, S., & Mohammadian, A. K. (2020). Toward safer highways, application of xgboost and shap for real-time accident detection and feature analysis. Accident Analysis & Prevention , 136 , 105405

    2020

  24. [32]

    J., & Tignor, S

    Payne, H. J., & Tignor, S. C. (1978). Freeway incident-detection algorithms based on decision trees with states. Transportation Research Record(682)

    1978

  25. [33]

    N., & Hall, F

    Persaud, B. N., & Hall, F. L. (1989). Catastrophe theory and patterns in 30-second freeway traffic data—implications for incident detection. Transportation Research Part A: General , 23 (2), 103–113

    1989

  26. [34]

    F., Noeimi, H., Sanwal, K., Rydzewski, D., Skabardonis, A., Varaiya, P., & Al-Deek, H

    Petty, K. F., Noeimi, H., Sanwal, K., Rydzewski, D., Skabardonis, A., Varaiya, P., & Al-Deek, H. (1996). The freeway service patrol evaluation project: Database support programs, and accessibility. Transportation Research Part C: Emerging Technologies, 4 (2), 71–85

    1996

  27. [35]

    W., & Koppelman, F

    Sermons, M. W., & Koppelman, F. S. (1996). Use of vehicle positioning data for arterial incident detection. Transportation Research Part C: Emerging Technologies, 4 (2), 87–96

    1996

  28. [36]

    S., & Schofer, J

    Sethi, V., Bhandari, N., Koppelman, F. S., & Schofer, J. L. (1995). Arterial incident detection using fixed detector and probe vehicle data. Transportation Research Part C: Emerging Technologies, 3 (2), 99–112

    1995

  29. [37]

    Sheu, J.-B. (2002). A stochastic optimal control approach to real-time, incident-responsive traffic signal control at isolated intersections. Transportation Science, 36 (4), 418–434

    2002

  30. [38]

    Singh, D., & Mohan, C. K. (2018). Deep spatio-temporal representation for detection of road accidents using stacked autoencoder. IEEE Transactions on Intelligent Transportation Systems , 20 (3), 879–887

    2018

  31. [39]

    J., & Chassiakos, A

    Stephanedes, Y. J., & Chassiakos, A. P. (1993). Application of filtering techniques for incident detection. Journal of transportation engineering , 119 (1), 13–26

    1993

  32. [40]

    Sutskever, I., Vinyals, O., & Le, Q. V. (2014). Sequence to sequence learning with neural networks. Advances in neural information processing systems , 27 . Taheri Sarteshnizi, I., Bagloee, S. A., Sarvi, M., & Nassir, N. (2024). Traffic anomaly detection: Exploiting temporal p...

    2014

  33. [41]

    Turner, B., Breen, J., & Howard, E. (2015). Road safety manual: a manual for practitioners and decision makers on implementing safe system infrastructure. https://roadsafety.piarc.org/ en/road-safety-management-safety-data/quality-and-under-reporting

    2015

  34. [42]

    Vapnik, V. (2013). The nature of statistical learning theory . Springer science & business media. 34

    2013

  35. [43]

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., . . . Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems , 30

    2017

  36. [44]

    J., & Zipser, D

    Williams, R. J., & Zipser, D. (1989). A learning algorithm for continually running fully recurrent neural networks. Neural computation, 1 (2), 270–280

    1989

  37. [45]

    Wirasinghe, S. C. (1978). Determination of traffic delays from shock-wave analysis. Transportation Research, 12 (5), 343–348

    1978

  38. [46]

    E., & Stoica, I

    Wu, Z., Jain, P., Wright, M., Mirhoseini, A., Gonzalez, J. E., & Stoica, I. (2021). Representing long-range context for graph neural networks with global attention. Advances in Neural Information Processing Systems , 34 , 13266–13279

    2021

  39. [47]

    Xie, Q., Guo, T., Chen, Y., Xiao, Y., Wang, X., & Zhao, B. Y. (2020). Deep graph convolutional networks for incident-driven traffic speed prediction. In Proceedings of the 29th acm international conference on information & knowledge management (pp. 1665–1674)

    2020

  40. [48]

    Yao, W., & Qian, S. (2020). Learning to recommend signal plans under incidents with real-time traffic prediction. Transportation Research Record, 2674 (6), 45-59. Retrieved from https://doi.org/10.1177/0361198120917668 DOI: 10.1177/0361198120917668

    2020 doi

  41. [49]

    Yuan, F., & Cheu, R. L. (2003). Incident detection using support vector machines. Transportation Research Part C: Emerging Technologies , 11 (3-4), 309–328

    2003

  42. [50]

    Zhang, Z., He, Q., Gao, J., & Ni, M. (2018). A deep learning approach for detecting traffic accidents from social media data. Transportation research part C: emerging technologies, 86 , 580–596

    2018

  43. [51]

    Zhu, W., Wu, J., Fu, T., Wang, J., Zhang, J., & Shangguan, Q. (2021). Dynamic prediction of traffic incident duration on urban expressways: A deep learning approach based on lstm and mlp. Journal of intelligent and connected vehicles , 4 (2), 80–91. 35 A Prediction/Detection E...

    2021

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.