Efficient streaming media caching policies help improve the request hit rate of Content Delivery Networks (CDNs), which leads to reduced service latency and enhanced service qual- ity. Existing research on caching policies primarily relies on the popularity of media content for decision-making. Me- dia servers face constraints in storage and computational re- sources, along with frequently changing service requests, mak- ing it challenging to determine the cached content and co- optimize caching and computational resources. We propose an innovative architecture for efficient CDN service caching policies to address this challenge. This architecture utilizes a hierarchical adaptive caching strategy (HACS) that deter- mines the computational resources to be allocated by pre- dicting the network resource consumption of each edge and cloud node. Next, we analyze the live and playback streams from each node using the Quantum Particle Swarm Opti- mization (QPSO) algorithm to extract features related to the playback streams. These features are used as inputs for pre- dicting potential hot content. Based on the prediction results and historical data, we dynamically adjust the caching pol- icy by monitoring resource consumption and user requests in real-time. Finally, experimental data demonstrate that, com- pared to existing schemes, our proposed architecture offers ∗Equally contributing authors. FIGURE 1 Scenarios of CDNs advantages in improving cache hit rates and reducing user latency.