Article: Research on the optimisation of communication efficiency based on adaptive improved federated learning Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.20 pp.19 - 40 Abstract: Aiming at the communication efficiency bottleneck in the internet of things and edge computing scenarios, this paper proposes a communication efficiency improvement scheme based on adaptive improved federated learning. By constructing an ARMA bandwidth prediction model enhanced by wavelet transform, the client network environment is predicted, and the improved Sketch compression algorithm is adopted to dynamically adapt to the real-time bandwidth conditions, thus the communication efficiency optimisation in the internet of things and edge computing scenarios is achieved. Experiments show that the accuracy of the proposed method researches 95%, the average uplink communication time is 0.5 seconds, and the communication efficiency exceeds 1.7. It provides key technical support for real-time federated learning deployment in 5G edge computing environment. Inderscience Publishers - linking academia, business and industry through research

Title: Research on the optimisation of communication efficiency based on adaptive improved federated learning

Authors: Xuefei Zhang; Yanli Zhao

Addresses: Department of Mechanical Engineering, Shanxi Engineering Vocational College, Taiyuan, 030000, China ' China Mobile Communication Group Design Institute Co., Ltd., Shanxi Branch, Taiyuan, 030000, China

Abstract: Aiming at the communication efficiency bottleneck in the internet of things and edge computing scenarios, this paper proposes a communication efficiency improvement scheme based on adaptive improved federated learning. By constructing an ARMA bandwidth prediction model enhanced by wavelet transform, the client network environment is predicted, and the improved Sketch compression algorithm is adopted to dynamically adapt to the real-time bandwidth conditions, thus the communication efficiency optimisation in the internet of things and edge computing scenarios is achieved. Experiments show that the accuracy of the proposed method researches 95%, the average uplink communication time is 0.5 seconds, and the communication efficiency exceeds 1.7. It provides key technical support for real-time federated learning deployment in 5G edge computing environment.

Keywords: federated learning; wavelet transform; ARMA; Sketch; communication efficiency.

DOI: 10.1504/IJICT.2025.146805

International Journal of Information and Communication Technology, 2025 Vol.26 No.20, pp.19 - 40

Received: 10 Dec 2024
Accepted: 26 Mar 2025
Published online: 18 Jun 2025 *

Title: Research on the optimisation of communication efficiency based on adaptive improved federated learning

Keep up-to-date