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

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