Distributed economic forecasting system and its application in rural revitalization

Abstract

AbstractCompared with the city, the rural area is large and the associated population disperses. Therefore, it needs to design a distributed scheme to collect and analysis the economic data. The federated learning can handle the economic data from different rural areas to protect primal economic data. This paper utilizes federated learning for economic forecasting which plays a very important role in rural revitalization. The proposed economic forecasting system consists of two parts: end nodes and cloud center. An end node is deployed in a village. The economic data of village is encrypted and villages only can communicate with other using encrypted data to avoid the leakage of primal data. The cloud center utilizes federated learning to jointly optimize the models in end nodes. During the processing of optimization, the transition between cloud center and end nodes is encrypted as well. The distributed economic forecasting system is evaluated on real data and achieves the expected results.