为减少边缘云雾协同网络的传输延迟,提高缓存设备的存储利用率和预测准确性,提出了一种基于用户兴趣与特征融合的数据预处理缓存策略(FUCS:A Fusion of User Interests and Features-based Data Pre-processing Caching Strategy).利用K-means算法对数据进行预处理,以缩小计算范围.设计了一个特征融合模块,并采用Multi-Head Self-attention来适应用户兴趣的变化规律.仿真试验结果表明,与传统的缓存策略相比,所提出的策略在总体缓存命中率上表现更优,并能显著降低数据的平均传输延迟.
FUCS:A Fusion of User Interests and Features-Based Data Preprocessing Caching Strategy
In order to reduce the transmission delay of edge cloud fog cooperative networks and improve the storage utilization and prediction accuracy of caching devices,a fusion of user interests and features-based data preprocessing caching strategy(FUCS)is proposed.The K-means algorithm is used to preprocess the data to narrow down the computational domain,a feature fusion module is designed,and the Multi-Head Self-attention is adopted to adapt to the changing patterns of user interests.The simulation results show that the proposed strategy performs better in overall cache hit rate and can significantly reduce the average transmission delay of data,compared with traditional buffering strategies.
国家科技期刊平台
NSTL
万方数据
