基于用户偏好预测的无人机部署和缓存策略

UAV deployment and caching scheme based on user preference prediction

针对蜂窝网络中的缓存问题,考虑用户内容请求的空间异构性及时间波动性,提出了一种基于单个用户内容偏好预测的蜂窝网中无人机位置部署及缓存内容部署方案。首先基于用户的历史上下文信息,利用文件相似性及用户相似性来预测每个用户的内容偏好特性,并使用一种基于线性回归的方法来预测用户未来发起内容请求时的位置和时间;然后根据预测的地理位置、请求时间和内容偏好,分别利用基于自组织映射神经网络(SOM)的聚类算法和基于凝聚嵌套(AGNES)的分簇算法设计无人机的部署位置,并根据相应的无人机位置设计内容部署方案。仿真结果表明,所提算法在缓存命中率和时延性能上均优于对比算法。对真实数据集的分析结果表明,不同的用户特征对内容偏好影响权重不等,因此需要对不同的用户特征赋予合理的权值。

In order to design an efficient edge caching policy considering spatial heterogeneity and temporal fluctuations of users’content requests,a proactive caching scheme was proposed with UAV’s deployment location design based on user preference prediction.Firstly,each user’s preference characteristics were predicted based on file similarity and user similarity,and the request time and user location were also predicted when a content request occurs.Thereafter,on the basis of the predicted geographical location,request time and user preference,each UAV’s deployment location and the corresponding content placement were determined by virtue of clustering method based on SOM and AGNES.Simulation results show that the proposed scheme outperforms other three comparison schemes in terms of hit ratio and transmission delay.Furthermore,the results also reveal that content preference is correlated with different user features by different weights.Accordingly,different impact weights should be matched with different user features.