This article proposes a mesh-based magneto-electric coupling network model for prediction and reduction of permanent magnet (PM) eddy current loss in a flux-switching permanent magnet machine. The key is to couple the electromagnetic inductions between the eddy current network and the equivalent magnetic network (EMN) in stator PMs dynamically. First, the eddy current networks of PMs are established with electric meshes. The electromotive forces of electric mesh are induced by considering double-side effects based on Maxwell's equations. Second, the stator meshing method is propose with PM segments, and combined in the EMN matching to the eddy current mesh. High efficiency and accuracy with detailed magnetic field distributions can be obtained. Then, the magnetic fields generated from EMN PM nodes are coupled in eddy current network to accurately predict the flow directions and density distributions. Finally, the original design and two other structures with rotor magnetic barriers are analyzed for reduction of eddy current losses. The effectiveness of the proposed model is verified by finite-element analyses and experimental results.