The present invention relates to a learning method of a model for generating point-of-interest (POI) data of a terminal through federated learning which comprises the steps of: receiving a short message service (SMS) for notifying a payment of a user of a terminal; inputting the SMS to a shop information extraction model and determining shop information from the output of the shop information extraction model; determining current location information on the terminal; caching POI data labeled with the store information in the location information; and learning a first common predicted model using the POI data. The first common predicted model is received through a server and the SMS includes text information indicating a name of a shop where the user made a payment.
展开▼