Page 237 - Big Data Analytics for Connected Vehicles and Smart Cities

218 Big Data Analytics for Connected Vehicles and Smart Cities Practical Applications and Concepts for Transportation Data Analytics 219

tation. The effort required in building such a bridge and bringing data scientists
up to speed, so to speak, on the characteristics of traffic data was underestimat-
ed. Similarly, the effort required to communicate data science to transportation
professionals was also underestimated. Other challenges related to the discovery
nature of the undertaking. The whole process of applying a discovery tool to
a data set, by its very nature means that things that were not perceived at the
beginning of the process can become important. It is also the case that one
discovery leads to another. This was certainly the case in this exercise, and con-
siderable effort was expended on the initial analysis of a significant extension
in the originally envisioned schedule for the work. The original work schedule
spanned a period of approximately two months, when, in fact, the actual work
spanned a period of more than 12 months.
Another challenge lay in the definition of TMC segments. These vary
in length from 500m to more than 2,000m. This limits the resolution of the
evaluation, as effects can only be analyzed over the length of the segment and
not within the segment. Since the work was completed, INRIX [1] has intro-
duced a new data set with shorter and more consistent segment lengths. This
supports a high-resolution of analysis and enables the possibility that this data
source could be used on arterials where the distances between intersections and
the variability in traffic speeds is greater. The technique developed during the
study will also be extremely valuable when used in conjunction with the con-
nected vehicle data. Connected vehicle data offers the possibility of second-by-
second speed profiles emanating from connected vehicles. This high-resolution
data set could be utilized with the techniques developed here to provide greater
insight into the driver behavior at the beginning, during, and at the end of a
bottleneck.
A significant discovery element of the project was a realization that the
analysis could form the basis for a new scientific approach to traffic engineer-
ing. As more data becomes available and as the accuracy and resolution of the
data grows, the principles revealed in this project could be applied to the adop-
tion of a scientific approach to traffic engineering.
This would be based on a detailed understanding of the variations in
traffic conditions and on the detailed effects of traffic management tools and
devices. With the advent of connected vehicle technology, which would enable
instantaneous vehicle speed, vehicle location, and vehicle ID to be gathered
on a large scale, it is likely that the ground will be prepared for a scientific ap-
proach to traffic engineering. It is our belief that connected vehicle data will
need to be incorporated into an integrated transportation data set that includes
crashes, incidents, road geometry, roadmaps, road signs, weather conditions
and other data on which discovery techniques can be applied to support a sci-
entific approach.