Third-party data screening method and device based on federal learning

Abstract

The invention relates to a third-party data screening method and device based on federal learning, electronic equipment, a computer readable medium and a computer program product. The method comprises the following steps: generating a target object set, wherein the training target set comprises identifiers of a plurality of target objects; each third-party data end generates corresponding model data according to the target object set; aggregating the plurality of model data in a longitudinal federated learning mode to generate a federated learning aggregation model; evaluating the performance of the federated learning aggregation model to generate a performance gain value; and according to the performance gain value, calculating a contribution score of each third-party data end so as to screen the third-party data. According to the method, the contribution score of each piece of third-party data can be obtained in a federated learning mode, so that dynamic screening of the third-party data is realized, and the model performance is improved and the interference of low-quality data is reduced on the premise of ensuring the data security.

Images