A Multilevel Graph Approach for Predicting Bicycle Usage in London Area

Abstract

Our cities are significantly changing their structure and organization. These changes have given rise to the need to introduce new services able to rationalize the activities present in such a complex context. Within this scenario, one of the most important services is related to transport management. A proper transport system management can significantly improve the overall quality of life of citizens, in terms of improving air quality, reducing road traffic and ensuring the public transport schedule. In addition to the traditional services and urban traffic management approaches, a significant contribution may come from the adoption of all those IT systems, which are part of the so-called Internet of things. According to this paradigm, it is possible to design an added value and pervasive services in order to assist the users involved in the system. Although this approach could be considered interesting and promising, it is necessary to introduce methodologies able to manage data coming from several heterogeneous sensors in order to process and propose coherent information. In this paper, we propose the using of three graphic approaches able to process the information coming from various sources in order to manage urban transport systems. The three models of representation, on which to conduct inference processes are context dimension tree, ontology and Bayes network. These three approaches allow the creation of inference processes, which represent the basis of value-added services to be offered to several users. The aim of this paper is to present a service that through a multilevel approach, which takes advantage of three models of graphic representation, is able to analyse data from various sensors in an urban area in order to predict the bicycle-sharing public service usage in the city of London. Through the intersection and analysis of data from cameras, weather and transport sensors, it will be possible to establish in which condition there will be an increase or decrease of bicycle rental in order to manage the service. The results obtained on data collected in real scenarios are very satisfying.