Urban mobility: Leveraging machine learning and data masses for the building of simulators

Urban mobility: Leveraging machine learning and data masses for the building of simulators

Keywords: Intelligent Public Transportation Systems, Data Science, Machine Learning, Digital Twins

Abstract: The so called data era we have entered in is accompanied by an explosion of data, both in variety and quantity. Public transportation is a data-intensive field, and related information systems are often supported by old technologies that struggle to keep up as the amount of data continually increases. This poses two problems. First, the massive data generated by the transportation network must be qualified and enriched with external data sources in order to be used for decision making. Second, in order to limit the number of tools and the complexity of maintenance, it is desirable to integrate data governance with decision support tools to allow non-expert operators to manipulate this data. Through four contributions leading to the proposal of a technical framework that integrates the past, present and future into a traditional information system containing a priori models, this thesis argues that the integration of various highly qualified datasets from the real world into a single spatio-temporal model provides a qualitative, efficient and low-cost mean of analysis, prediction and strategic decision support for bus networks while depreciating the use of data management systems in a non integrated multi-tool data management systems?