Based on the cold wave activity data of the winter half year from 1965-2015 and NCEP/NCAR reanalysis data, the study applies finite-mixed-model algorithm based clustering analysis to identify the cold wave paths affecting China as four clusters. The activity variation trends and circulation characteristics of the four clusters of cold waves, with different paths are analyzed. The results prove that the frequency, intensity, seasonal distribution and inter annual variation of cold waves, with different paths are significantly different. Cluster-1 cold waves which enter from the west and then turn its direction appear most frequently and have the largest tendency of year-by-year decrease, the cluster-3 code waves, which enter from the west are with the highest intensity, cluster-4 cold, waves which are from the west appear frequently in spring and the variation trend is gradual. During the cold outburst, temperature-pressure field configuration, transition of wind field, distribution of high-level advection of vorticity and temperature advection and surface cold high pressure are closely related with selection of cold wave path. The low trough, negative vorticity area and cold advection area of cluster-2 code waves which are from the north are distributed eastward, while those of cluster-3 code waves are distributed westward.