The invention discloses a subway crowd behavior label identification method based on a base station position, and the method comprises the following steps: S1, obtaining the user data of a subway operator base station, processing the user data, and obtaining the position of a cellphone user time dimension; S2, building a subway crowd recognition model: enabling the operator base stations to correspond to subway stations, building a subway region configuration table, and recognizing subway crowds; and S3, establishing a crowd label model, and performing statistics to obtain the number of inbound people, the number of outbound people and the number of exchanged people of the subway station in a fixed time interval. According to the method, the metro crowd behavior label is obtained by combining operator user signaling data, base station positions and a modeling algorithm. In combination with a previous subway crowd monitoring means, subway crowd monitoring is completed through another scene, and automation and precision are achieved.