<![CDATA[ In modern computing systems, caching is an essential technique for improving data access efficiency. However, traditional caching algorithms such as LRU and LFU have limitations in handling complex data dynamics. This research proposes an adaptive caching algorithm that integrates semantic context and access time profiles to enhance system performance. The approach utilizes a semantic embedding model based on Sentence-BERT and temporal analysis of user access patterns. Testing was conducted through simulations using real and synthetic datasets, and compared with conventional caching methods such as KVShare and LRU. Evaluation results show that the proposed algorithm is capable of increasing the cache hit rate by more than 83%, reducing average latency to around 61 ms, and maintaining resource usage efficiency. In addition, the algorithm demonstrates strong adaptability to dynamic access pattern changes and responsiveness to semantic parameter adjustments. Thus, the integration of semantic context and temporal information provides significant contributions to optimizing cache management. This algorithm has potential applications in edge computing systems, LLM services, and cloud-based platforms. Suggestions for future research include implementation in real-world environments, application of predictive machine learning models, and dynamic exploration of adaptive parameters. ]]>
Copyrights © 2025
