Detecting interpersonal relationships in large-scale railway trip data

With increases in the amount of human trajectory data, interest in explaining or predicting human mobility is growing. Owing to the difficulty of associating mobility data with interpersonal relationship data, previous studies on the link between interpersonal relationships and mobility are limited to the specific activities of particular users. In this paper, we propose a method for detecting interpersonal relationships from mobility data, while distinguishing these relationships from those of familiar strangers such as commuters. In the method, persons who take diverse variations within the same activities are recognized as a pair. From IC card data covering the daily mobility of six million people over three years, we detected millions of frequently co-located pairs. Under certain conditions, most of the detected pairs are confirmed as not being familiar strangers, but rather to have an interpersonal relationship. Next, we analyzed the detected pairs and found that the density of the relationships between groups was divided by gender and age and was found to be asymmetric by gender. For example, an elderly male person is not likely to take trips as a pair with a same-gender elderly person, and this result is data-based evidence for the isolation of retired men. In addition, group trips are confirmed to have an extraordinal character and sometimes converge spatiotemporally. These findings indicate that interpersonal relationship is a strong factor to determine their mobility and group observation is potentially useful for event detection.