Regional-union based federated learning for wireless traffic prediction in 5G-Advanced/6G network
Jiahao Nan, Ming Ai, Aijuan Liu, Xiaoyan Duan · 2022
With the development of artificial intelligence, wireless traffic prediction plays an important role in the intelligence of wireless communication, such as load balancing and energy saving of base stations. Most of the existing prediction methods adopt the centralized training method, which requires the data to be centralized, which will cause data security and privacy issues. We introduce a framework named region-union based federated learning(FedRU) to solve this problem. The central node trains the shared model according to the regional model, and each region trains its own regional model. The shared model not only takes into account the global user characteristics, but also ensures the regional data characteristics to the greatest extent. We verify performance on real-world datasets through comparative experiments, and the results show that after only 27 rounds of communication, the accuracy rate can reach 0.80, and finally after 50 rounds of communication, the accuracy rate can reach 0.86.