Caracterisation des lieux d'activites a partir de donnees de cartes a puce de transport collectif.

摘要

The use of smart-card in transit system has developed a lot since the end of the 90's. Their first goal is to collect fare automatically and to check instantly the ticket validity. These systems produce a huge amount of data that can be used to analyze a transit network. Actually, with the databases stemming from such systems, we can have a full and continuous picture of travelers' behaviors. These behaviors are conditioned by the geography of the area and the structure of the network. For instance, people use public transportation mostly for commuting. That's why city centers attract most users during the morning peak hours. These variations in space and time will be the core of this study.;In the first part, we are going to present a literature review about variability analysis and models of transportation users' behaviors, smart-card systems and their interests and finally, data mining and its application on geospatial databases.;In a second part, we are going to present the STO public transportation network and to explain the methodology used to characterize activity places. The method is divided in two main phases: (1) On one hand, we are going to study bus stops' behaviors by clustering the bus stop population in a small number of homogeneous groups with data mining algorithms. During this phase, one of the major issues we face is the determination of an appropriate number of groups. Thus, after having discussed and identified a appropriate number of groups, we are going to study the stops variability within these groups in order to decrease the data noise. (2) On the other hand, we are going to create a geographic information system (GIS) of the STO public transportation network. This GIS locates all major activity places in the Ottawa region and all the STO bus stops. Then, we are going to define proximity relations between GIS elements to, finally, find a characterization for the activity places.;The end of this second part is presenting the databases used to undertake the studies and is describing the main facts of the network (number of boardings per hour per day, origin-destination...).;We are going to analyze the influence of major activity locations on a public transportation network using smart-card data from the Outaouais Transit Company (Societe des transports de l'Outaouais, STO). That is to say, we are going to see how users who board near a main trip generator usually behave. The volume of data in the transportation database is so huge (more than 800,000 data a month) that we are going to use data mining techniques to extract knowledge from them.;The third part is applying the methodology and extracting the results. Actually, we are going to cluster our bus stops population in nine typical behavior groups. The study of the belonging group variability will show us that groups are composed of a core of non-variable stops and a cloud of very unpredictable elements. It will permit us to decrease the noise in the data. By analyzing the proportion of each type of users (Adult, elderly, student) in each group, we will see that there is a majority of students on stops mostly used between 9:00 am and 16:00 pm and that they tend to make shorter trips. This analyze also confirms that users living far from the city center use transportation earlier in the morning and they tend to do longer trips. And finally, the activity places characterization reveals us that car parks create boardings mostly between 7:00 and 8:00 in the morning, the city center between 16:00 and 17:00, hospitals and shopping centers create a regular use during all the day.