With the increasing number of large special winter outdoor events, it is necessary to analyze the characteristics of passenger flow demand and conduct data mining. Based on the passenger flow demand survey, the characteristics and intentions of passenger demand are analyzed, it is concluded that the acceptable reasonable walking and queuing time for people is within 20 minutes, and the outdoor competition time is within 3 hours. And heating and catering service facilities should be provided during the competition, and a viewing area near the stadium mainly for walking and watching should be set up to meet the needs of spectators. Then the user portraits of the three groups are established by using the second-order clustering and hierarchical clustering method respectively. Through the analysis of 3D map of the clustering effect, silhouette coefficient and clustering user profile characteristics, it is concluded that the second-order clustering effect is better. Establish three types of tag cloud maps, local group viewing type: local middle-aged and elderly people are mostly local, and the time for walking, queuing, and watching the competition is short, it is recommended to reasonably plan the drop-off area for vehicles of middle-aged and elderly groups, and orderly organize the arrival of passengers. Companion free viewing type: groups have higher time thresholds for walking, queuing and viewing time. It is recommended to set up entertainment and rest points to attract groups to play. Sightseeing type for outland youths; the group is unfamiliar with the area, the time for walking, queuing and watching the game is medium, it is recommended to strengthen the guidance of transportation facilities inside and outside the competition area, and should rationally plan public transportation stations, parking lots and other transportation service facilities.