• Activity-Based
• Attendance Prediction
• Ride Matching Algorithms
• Ride-sharing
• Social Media

People living in highly-populated cities increasingly suffer an impoverishment of their quality of life due to pollution and traffic congestion problems caused by the huge number of circulating vehicles. Indeed, the reduction the number of circulating vehicles is one of the most difficult challenges in large metropolitan areas. This PhD thesis proposes a research contribution with the final objective of reducing travelling vehicles. This is done towards two different directions: on the one hand, we aim to improve the efficacy of ride sharing systems, creating a larger number of ride possibilities based on the passengers destination activities; on the other hand, we propose a social media analysis method, based on machine learning, to identify transportation demand to an event. Concerning the first research direction, we investigate a novel approach to boost ride sharing opportunities based, not only on fixed destinations, but also on alternative destinations while preserving the intended activity of the user. We observe that in many cases the activity motivating the use of a private car (e.g., going to a shopping mall) can be performed at many different locations (e.g. all the shopping malls in a given area). Our assumption is that, when there is the possibility of sharing a ride, people may accept visiting an alternative destination to fulfill their needs. Based on this idea, We thus propose Activity-Based Ride Matching (ABRM), an algorithm aimed at matching ride requests with ride offers to alternative destinations where the intended activity can still be performed. By analyzing two large mobility datasets, we found that with our approach there is an increase up to 54.69% in ride-sharing opportunities compared to a traditional fixed-destination-oriented approach. For the second research contribution, we focus on the analysis of social media for inferring the transportation demands for large events such as music festivals and sports games. In this context, we investigate the novel problem of exploiting the content of non-geotagged posts to infer users’ attendance of large events. We identified three temporal periods: before, during and after an event. We detail the features used to train the event attendance classifiers on the three temporal periods and report on experiments conducted on two large music festivals in the UK. Our classifiers attained a very high accuracy, with the highest result observed for Creamfields festival (∼91% accuracy to classify users that will participate in the event). Furthermore, we proposed an example of application of our methodology in event-related transportation. This proposed application aims to evaluate the geographic areas with a higher potential demand for transportation services to an event.