Database Reference
In-Depth Information
other, and hopefully evolve into next-generation traffic modeling systems able
to better capture the dynamics of real human mobility. The following sections
try to trace some connections between the two fields, providing examples of a
mobility data mining approach to deal with some basic questions that naturally
arise when dealing with traffic understanding and modeling. Moreover, the
remaining part of the chapter will provide a series of real analytical scenarios
enabled by the methods and techniques presented in the previous chapters of this
topic.
10.2 Data-Driven Traffic Models
Mobility phenomena are sensed by means of several data collections and mon-
itoring. For example, traditional transportation methods use inductive loops,
cameras, sensors, and counters to measure specific arc roads of the network. All
these observations are merged and integrated within existing models in order to
refine and fit the model parameters. Thus, the integration of the mobility models
extracted from real mobility data is crucial. There is a strong need for an accurate
mobility demand evaluation that calls for a data-driven approach to obtain better
estimations of mobility phenomena. In this chapter we will show how to cope
with a set of problems that provide the analyst with a particular view on specific
mobility behaviors. At the base of such a process there is a large preprocessing
step with the duty of integrating and merging different data sources. For the
objective of this chapter we assume that this step has already been performed
and all the data are available for the analysis in the correct format. We show
how to master the complexity of the mobility knowledge discovery process by
means of an organic analytical framework centered on the concept of
trajectory
.
In particular, we show how the semantic deficiency of big mobility data can be
bridged by their size and precision. To this purpose, we describe the key results
obtained in a large-scale experiment conducted with the mobility analysts of the
cities of Milan and Pisa, on the basis of real life GPS tracks sensed from tens of
thousands of private cars. We show how it is possible to find answers to chal-
lenging analytical questions about mobility behavior, which are not supported
by the current generation of commercial systems, such as:
1. What are the most popular itineraries followed from the origin to the desti-
nation of people's travels? What are the routes, timing, and volume for each
such itinerary?
2. How do people leave the city toward suburban areas (or vice versa)? What is
the spatio-temporal distribution of such trips?
3. How can we understand the accessibility to key mobility attractors, such as
large facilities, railway stations, or airports? How do people behave when
approaching an attractor?
