A cloud-edge-end collaborative intelligent caching method based on incremental federated learning algorithms.

In a cloud-edge-end collaborative system, data generated by terminal devices often contains users' sensitive information and is constantly generated and changing, leading to potential data privacy leaks in caches. Additionally, due to the inability to promptly capture these dynamic changes and the failure to consider the actual capabilities of nodes, caching strategies become outdated, resulting in reduced cache hit rates and cache imbalance issues. Therefore, this study proposes a cloud-edge-end collaborative intelligent caching method based on an incremental federated learning algorithm. First, the federated learning algorithm is used to aggregate data from terminal devices to the cloud, enabling collaborative data processing while protecting data privacy. Second, incremental learning methods are employed to continuously update terminal data, with the updated data aggregated to the cloud, thereby enabling real-time tracking of data trends and allowing cache strategies to rapidly adapt to dynamic changes in terminal data. Finally, considering the actual capabilities of nodes, the popularity of aggregated data and the weights of edge and terminal nodes are calculated. Data is cached in edge and terminal nodes in descending order of popularity and weight. When cache space is insufficient, data replacement is performed based on the importance of data within nodes, thereby completing intelligent data caching. Experimental results demonstrate that this method achieves good performance in data update aggregation, with high data caching balance and cache hit rates.