<![CDATA[The objective of this study is to develop a trusted AI-based framework for predicting software-defined networking (SDN) controller load and optimizing fog/edge microservice orchestration to reduce end-to-end latency in dense 5G scenarios. The proposed approach integrates user-aware spatial clustering with evolutionary resource selection to maintain stable quality of service (QoS) under high mobility and traffic variability. In the analysis stage, k-means clustering partitions users into spatial sectors and identifies sector centroids. Particle swarm optimization (PSO) is then applied to fog-node selection, resource sizing, and adaptive microservice placement and migration. To enhance system resilience, a recurrent neural network (RNN) is employed to forecast SDN controller load using correlation-informed features extracted from service-channel dynamics. Numerical experiments on heterogeneous fog-node topologies indicate that the framework reduces microservice execution time by 69% relative to baseline placement strategies under identical load conditions, while controller-load prediction attains an RMSE of 0.00387. These findings confirm the effectiveness of both the latency-reduction mechanism and the controller-load estimation workflow. The novelty of this work lies in the unified optimization of microservice placement, migration, and SDN controller-load anticipation within a single reproducible architecture, extending existing fog and edge orchestration approaches that typically address these components as independent subproblems.]]>
