基于深度学习的智能移动边缘网络缓存

针对移动边缘网络缓存问题，提出把计算资源推送到网络边缘，使边缘接入热点能有数据分析能力，构建基于深度学习的深度缓存策略，进一步提升缓存效率。在边缘接入热点处构建基于长短期记忆神经网络模型的缓存内容文件流行度预测系统，通过分析本地数据给出内容文件流行度预测。把内容文件流行度预测系统整合到移动边缘网络缓存系统中最大化缓存命中率，提出深度缓存策略，大大提升移动边缘网络缓存性能。在真实视频数据集上进行测试，实验结果表明：提出的内容流行度预测系统的准确度高于现有最优方法；提出的深度缓存策略与传统的缓存算法相比，在相同的缓存命中率指标下大约仅需一半的缓存存储空间。

Intelligent mobile edge network caching based on deep learning

In view of mobile edge network caching problem, computation resources are further pushed to the network edge to enable data analysis and to build deep learning-based caching strategy at access points, thereby boosting caching gain. The long short term memory (LSTM)-based neural network is proposed to predict the future content popularity by analysing the local data, which is further used to optimize content replacement for the cache hit rate maximization and construct deep caching strategy. Real-world dataset is used to validate the effectiveness of the proposed deep caching strategy. Numerical results demonstrate that our content popularity prediction method outperforms the state-of-art prediction method. Compared with traditional methods, the caching system needs only approximately half storage space to achieve the same cache hit rate.