Resumen
Carsharing is a new alternative transport mode which is a good solution to balance travellers? requirements between private vehicles and public transit. Understanding carsharing use patterns benefits the planning and design of a better service. The criteria for use behavior classification is imprecise and short of data support. This paper introduced clustering method to analyze carsharing use patterns based on carsharing rental data provided by Chefenxiang in Hangzhou, China. A set of descriptive variables were selected to describe carsharing use behaviors and member travel features. By calculating the use count, use location count, use time length and use time interval of carsharing members, five clusters were finally formed, named as short-term frequent heavy use pattern, short-term frequent light use pattern, long-term frequent light use pattern, long-term occasional light use pattern and long-term frequent heavy use pattern. More detailed information such as start time and end time distribution, location frequency and cluster size compositions and spatio-temporal distribution in workdays and nonworkdays was discussed. Based on the spatio-temporal analysis on these five clusters, it was concluded that long-term frequent heavy members were inferred as stable users for commute or business purpose. Carsharing in China attracted college students and people around office buildings and business areas. Most of the travels made by carsharing were short or middle distanced and temporary. Using clustering method to mine use patterns was valid without priori knowledge about categories of use behaviors. The characteristics of Chinese carsharing use patterns could help make questionnaire options more scientific and policies for particular groups of members more reasonable.