MOBILITY ANALYSIS AND PROFILING FOR SMART MOBILITY SERVICES: A BIG DATA DRIVEN APPROACH. An Integration of Data Science and Travel Behaviour Analytics

Smart mobility proved to be an important but challenging component of the smart
cities paradigm. The increased urbanization and the advent of sharing economy require
a complete digitalisation of the way travellers interact with the mobility services.
New sharing mobility services and smart transportation models are emerging as partial
solutions for solving some tra c problems, improve the resource e ciency and reduce
the environmental impact. The high connectivity between travellers and the sharing
services generates enormous quantity of data which can reveal valuable knowledge and
help understanding complex travel behaviour. Advances in data science, embedded
computing, sensing systems, and arti cial intelligence technologies make the development
of a new generation of intelligent recommendation systems possible. These
systems have the potential to act as intelligent transportation advisors that can o er
recommendations for an e cient usage of the sharing services and in
uence the travel
behaviour towards a more sustainable mobility. However, their methodological and
technological requirements will far exceed the capabilities of today's smart mobility
systems.
This dissertation presents a new data-driven approach for mobility analysis and travel
behaviour pro ling for smart mobility services. The main objective of this thesis is
to investigate how the latest technologies from data science can contribute to the
development of the next generation of mobility recommendation systems.
Therefore, the main contribution of this thesis is the development of new methodologies
and tools for mobility analysis that aim at combining the domain of transportation
engineering with the domain of data science. The addressed challenges are derived from
speci c open issues and problems in the current state of the art from the smart mobility
domain. First, an intelligent recommendation system for sharing services needs a
general metric which can assess if a group of users are compatible for speci c sharing
solutions. For this problem, this thesis presents a data driven indicator for collaborative
mobility that can give an indication whether it is economically bene cial for a group
of users to share the ride, a vehicle or a parking space. Secondly, the complex sharing
mobility scenarios involve a high number of users and big data that must be handled by
capable modelling frameworks and data analytic platforms. To tackle this problem, a
suitable meta model for the transportation domain is created, using the state of the art
multi-dimensional graph data models, technologies and analytic frameworks. Thirdly,
the sharing mobility paradigm needs an user-centric approach for dynamic extraction
of travel habits and mobility patterns. To address this challenge, this dissertation
proposes a method capable of dynamically pro ling users and the visited locations in
order to extract knowledge (mobility patterns and habits) from raw data that can be
used for the implementation of shared mobility solutions. Fourthly, the entire process of
data collection and extraction of the knowledge should be done with near no interaction
from user side. To tackle this issue, this thesis presents practical applications such
as classi cation of visited locations and learning of users' travel habits and mobility
patterns using historical and external contextual data.