Many-Objective Optimization Based-Content Popularity Prediction for Cache-Assisted Cloud-Edge-End Collaborative IoT Networks

With the advancement of mobile communication technology, there has been a marked increase in the demand for personalized and ubiquitous Internet-of-Things (IoT) services, raising the expectations for network quality of service (QoS) and quality of experience (QoE). Existing popularity prediction-based content caching policies improve QoS and QoE by pre-caching contents at the network edge, but jointly optimizing multiple network metrics remains a challenge. To address this challenge, we propose a many-objective optimization-based popularity prediction for cooperative caching (MaOPPC-Caching) framework for cloud-edge-end collaborative IoT networks. This framework simultaneously optimizes prediction accuracy, delay, offloaded traffic, and load balance. We integrate three prediction algorithms to forecast content popularity and present a horizontal and vertical collaborative caching decision strategy to generate caching forms based on the predicted results. Then, the many-objective evolutionary algorithm (MaOEA) is employed to optimize the combined proportions to take full advantage of hidden preferences and popularity characteristics of both users and items. To promote the convergence of the framework, we present a knowledge mining-based MaOEA (KMaOEA) to incorporate knowledge mining into the optimization process. Simulation results show that the proposed MaOPPC-Caching framework outperforms existing prediction algorithms in terms of four evaluation indicators. Furthermore, KMaOEA shows a significant advantage over NSGA-III in load balance, as indicated by a Mann-Whitney rank sum test with a p-value of 0.040.