Abstract: It is a prerequisite to optimize the operation of a shared bicycle system by collecting spatial-temporal characteristics of it.A web crawler technology is used to obtain data of stations.After that, indices like activity score (AS) are defined to measure operation of the system.In order to identify a valid cluster algorithm for computing AS, several methods are compared by using Dunn index and Davies-Bouldin index.Global and local Moran's statistics are introduced to analyze spatial-temporal characteristics of stations usage.In a case study of a shared bicycle system in Suzhou, stations are clustered into four types according to the changing regularity of AS.It is found that activity of one station has positive spatial correlation with 13 stations during rush hours in evening and 20 stations during rush hours in morning.It is speculated that average riding distance are 1.7 km and 2.2 km, respectively.Although majority of the normalized available bicycles (NABs) at peak times are spatially distributed at random, this study still finds that stations with high NAB are gathered in several different areas.Stations with low NABs are centrally located in a larger area in the city.The results also clearly reveal existing problems of coordination between stations in the system.