An AI-Based Temporal-Structural Fusion Framework for Robust Backend Load Prediction in Cloud-Native Environments

This paper proposes a graph-structured temporal dynamic learning model to address the challenges of backend load prediction in cloud computing and microservice environments, including dynamic topology changes, complex dependency structures, and multi-source heterogeneous monitoring data. The model constructs a time-varying service dependency graph to adaptively model structural relationships among nodes and integrates a temporal encoding mechanism to capture multi-scale temporal features, achieving joint representation of load characteristics in both spatial and temporal dimensions. It consists of four main modules: dynamic graph construction, graph convolutional feature extraction, temporal encoding, and spatiotemporal fusion. By jointly optimizing structural dependencies and temporal evolution, the model enables efficient interaction and representation of multidimensional features. Based on real backend system monitoring data, sensitivity experiments were conducted on hyperparameters such as the number of graph convolution layers, attention heads, adjacency matrix sparsity, and node count, systematically evaluating performance variations under MSE, MAE, MAPE, and RMSE metrics. Experimental results demonstrate that the proposed model significantly outperforms mainstream methods in both prediction accuracy and stability, maintaining strong robustness and generalization capability under complex topological conditions. The study confirms that graph-structured temporal dynamic learning effectively captures the structural evolution and temporal dynamics of backend systems, providing reliable technical support for intelligent load management and resource optimization in cloud platforms.