Discovering Regularity in Mobility Patterns to Identify Predictable Aggregate Supply for Ridesharing

Abstract

Heterogeneous data collected by smartphone sensors offer new opportunities to study a person’s mobility behavior. The mobility patterns extracted from the travel histories found in these data enable agents residing in mobile devices to model transitions between visited locations, so that upcoming trips can be predicted after observing a set of events, and assistance can be planned in advance. When several agents cooperate, the forecasted trips made by multiple users can provide a potential supply for shared mobility systems such as dynamic ridesharing. These trips must be sufficiently regular and frequent to be reliably announced as shareable trips. This paper describes a methodology to identify a predictable aggregate supply for ridesharing via mobility patterns discovered in users’ travel histories. The methodology empirically quantifies measures like the regularity and frequency of these patterns on a dataset consisting of 967 users scattered across different geographical areas. The sample exhibits high heterogeneity with respect to these measures (hence, of predictability, regardless of the prediction method). This paper shows how frequency of trip patterns decreases, while regularity increases, when additional dimensions such as departure times are added to the analysis. It was concluded that the traveler flexibility with regard to accepting less regular trips is vital to discover a larger supply. These results provide insights to develop future applications able to take advantage of this approach, to increase ridesharing rates, allowing a critical mass to be more easily attained.