A Question of Trust: Statistical Characterization of Long-Term Traffic Estimations for their Improved Actionability

Laña, Ibai; Villar-Rodríguez, Esther; Etxegarai, Urtats; Oregi, Izaskun; Ser, Javier Del

doi:10.1109/itsc.2019.8916850

Cited by 4 publications

(6 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Long-term traffic estimation models seek different traffic patterns (e.g. typical daily traffic profiles), and decide which of these patterns suits best the traffic behavior of the selected road for the date under choice [211]. The chosen pattern among all those elicited by the model becomes the prediction for the entire interval.…”

Section: Critical Analysismentioning

confidence: 99%

“…From a strategic point of view, confidence estimation in travel demand prediction has a solid research background [238], [239], [240], [241], [242], which helps design and scale properly road infrastructure. Confidence for longterm congestion predictions have also relevant contributions [211], [243]. However, there are no remarkable contributions on this matter for short-term traffic forecasting.…”

Section: Challenges and Opportunitiesmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning for Road Traffic Forecasting: Does it Make a Difference?

Manibardo¹,

Laña²,

Ser³

2020

Preprint

View full text Add to dashboard Cite

Deep Learning methods have been proven to be flexible to model complex phenomena. This has also been the case of Intelligent Transportation Systems (ITS), in which several areas such as vehicular perception and traffic analysis have widely embraced Deep Learning as a core modeling technology. Particularly in short-term traffic forecasting, the capability of Deep Learning to deliver good results has generated a prevalent inertia towards using Deep Learning models, without examining in depth their benefits and downsides. This paper focuses on critically analyzing the state of the art in what refers to the use of Deep Learning for this particular ITS research area. To this end, we elaborate on the findings distilled from a review of publications from recent years, based on two taxonomic criteria. A posterior critical analysis is held to formulate questions and trigger a necessary debate about the issues of Deep Learning for traffic forecasting. The study is completed with a benchmark of diverse short-term traffic forecasting methods over traffic datasets of different nature, aimed to cover a wide spectrum of possible scenarios. Our experimentation reveals that Deep Learning could not be the best modeling technique for every case, which unveils some caveats unconsidered to date that should be addressed by the community in prospective studies. These insights reveal new challenges and research opportunities in road traffic forecasting, which are enumerated and discussed thoroughly, with the intention of inspiring and guiding future research efforts in this field.

show abstract

Section: Critical Analysismentioning

confidence: 99%

Section: Challenges and Opportunitiesmentioning

confidence: 99%

Deep Learning for Road Traffic Forecasting: Does it Make a Difference?

Manibardo¹,

Laña²,

Ser³

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Traffic volumes, typically examined in both research and practice using 24h time series, are well known to show systematic volume variations during a day and between days. Time of day is an important predictor for network usage with 24h time series collected at a single point showing a regular M-shaped curve (Crawford et al, 2017;Laña et al, 2019;Weijermars & Van Berkum, 2005). Considering the day-to-day variability, traffic volumes deviate systematically between days (Ma et al, 2021;Stathopoulos & Karlaftis, 2001), and show considerable seasonable variation (Coogan et al, 2017;Thomas et al, 2008).…”

Section: Modeling Traffic Variationsmentioning

confidence: 99%

“…It is recognized in intersection design manuals (CROW, 2006) that traffic shows random behavior and that therefore delays differ from person to person, and that the performance of an intersection changes over time. Although almost all major cities adopted urban traffic management systems in various forms (see, e.g., Hamilton et al, 2013;Nellore & Hancke, 2016), these systems rarely take the accompanying uncertainties explicitly into account (Tettamanti et al, 2011) although, for example, a quantification of the (prediction) uncertainty helps managers to value forecasts (Laña et al, 2019). In the relative comfort of model predictive control, where a model is assumed to mimic the real-world environment, it was shown that taking uncertainties into consideration improves control decisions (Hu & Hellendoorn, 2013;Tettamanti et al, 2011).…”

Section: Traffic Variations and Urban Traffic Managementmentioning

confidence: 99%

“…Although useful for longer-term predictions, shorter-term fluctuations that cover less than 24h, e.g., events with varying starting times, are not incorporated. 24h patterns are also used by Laña et al (2019), where several clusters of time series are constructed using a neural network. Based on extracted features, a 24h prediction is constructed by selecting the mean time series of the corresponding cluster.…”

Section: Literature Overview and Research Contributionmentioning

confidence: 99%

See 1 more Smart Citation