Abstract: Taxis running in urban space generate a lot of OD data. The time-series of data presents periodic characteristics and objectively reflects people's travel behavior patterns. The spectral clustering algorithm is used to explore the travel characteristics and similarities of spatial units in the Fifth Ring Road area of Beijing. Because the temporal-spatial behavior characteristics are affected by the neighborhood and functional zoning, the study adds the neighborhood factor and the functional area factor to improve the time series similarity measurement method, and realizes the space and functional area expansion of the time series spectrum clustering algorithm. Furthermore, it can increase people's understanding of travel behavior characteristics under different time and space conditions, so as to provide different services for different spatial units, such as the frequency of bus in different time periods, the dynamic adjustment of the business hours of malls, the real-time transformation of taxi waiting points in different time and space environments, and the regulation and optimization of service guarantees in different regions, etc. The research helps to reduce urban energy consumption, makes more rational use of resources, and helps residents to realize smart life.

Key words: taxi data, the Fifth Ring Road area of Beijing, travel behavior, time series, spectral clustering