Federal learning supported real estate financial risk privacy calculation method and platform

The invention relates to a method and a platform for privacy calculation of real estate financial risks supported by federated learning. The method comprises the following steps: acquiring original data acquired by each participant on an information chain of a real estate financial network; identifying and determining a data source entity and privacy features, and performing homomorphic encryption on the privacy features to form a privacy protection data set of each information node; co-training the privacy protection data set through federated learning to obtain a real estate mortgage loan risk assessment model, and assessing the extracted privacy features to obtain a borrower credit risk result and a collateral risk assessment result; and calculating the probability of default of the real estate mortgage loan and the loss amount of the financial institution after the default, and determining a real estate mortgage financial risk result. According to the method, multiple participants are allowed to jointly t