<p>Efficient traffic management in telecommunication networks is essential for maintaining high-quality service and user experience. Multiprotocol Label Switching (MPLS) is widely used for traffic routing, where load balancing is crucial in optimizing resource utilization. Existing methods failed to achieve dynamic optimization in resource provisioning and traffic load balancing. As a result, the network gets congested, and the performance deteriorates. This manuscript proposes a new GGCNN that combines the Hybrid Golden Jackal and Chimp Optimization Algorithm (Hyb-GJ-COA) in addressing these issues. The first phase of the methodology is guided nonlinear anisotropic filtering which enhances the quality of input data by preprocessing network traffic flow data. The GGCNN model is an efficient resource provisioning and traffic load balancing platform, whereas the Hyb-GJ-COA improves the performance of the GGCNN model by optimizing its parameters. The proposed deep learning and nature-inspired optimization methodology will go a long way in reducing resource provisioning costs and traffic load balancing costs while enhancing metrics of Quality of Service by throughput, overhead, and Quality of Experience for end-users. Extensive evaluations using the Python platform show that the proposed GGCNN-Hyb-GJ-COA method outperforms the approaches in place. The proposed method has achieved 7.52%, 13.5%, and 11.06% improved throughput; 32.75%, 30.97%, and 29.87% overhead reductions; 24.43%, 27.74%, and 23.75% reductions in TLBC; and 4%, 29.04%, and 13% less expensive resource provisioning against the existing scheme like Pareto-based Modified Local Global Particle Swarm Optimization (PMLG-PSO), Bandwidth Allocating Algorithm in MPLS, and Path Computation Algorithms in Generalized MPLS Networks.</p>