Many mobile applications such as Strava or Mapmyride allow cyclists to collect detailed GPS traces of their trips for health or route sharing purposes. However, cycling GPS traces also have a lot of potential from an urban planning perspective. In this paper, we focus on two important issues to characterize urban cyclist behavior: trip purpose and route choice. Cycling trip purpose has been typically analyzed using survey data. Here, we present a method to automatically infer the purpose of a cycling trip using cyclists' personal data, GPS traces and a variety of built-in and social environment features extracted from open datasets characterizing the streets cycled. We evaluate the proposed method using GPS traces from over 7, 000 cycling routes in the city of Philadelphia and report F1 scores of up to 86% when four trip purposes are considered. On the other hand, we also present a novel statistical method to identify the role that certain variables characterizing the built-in and social environment play in the selection of a specific cycling route. Our results show that cyclists in Philadelphia tend to favor routes with green areas, safety and centrality.