Abstract: In transparent computing, the server stores and manages the operating system, application and personalized data, and processes users’ resource request from transparent networks. So the server is the bottleneck of system performance. User behavior analysis and prediction has received extensive attention in the field of network computing and social networks. However, there is currently no related work focused on the user behavior characteristics mining under the transparent computing. Based on information entropy and cubic exponential smoothing, the user behavior of transparent computing is analyzed and predicted, to provide effective basis for more efficient cache policies. Firstly, the user requirements are analyzed based on information entropy, and then the exponential smoothing algorithm is used to predict the access frequency of  blocks in the near future. The experiments on the real data are made to test the effectiveness of prediction.

Key words: transparent computing, user access behavior, information entropy, exponential smoothing, behavior prediction

摘要： 在透明计算中，服务端存储并管理着所有用户所需的操作系统、应用软件和个性化数据，并高效处理来自透明网络的用户资源请求服务。因此，服务端是透明计算系统性能的瓶颈。为制定更高效的缓存策略提供有效的依据，基于信息熵和三次指数平滑对透明计算用户行为特征进行分析和预测。首先基于信息熵策略分析用户访问行为特征，进而利用指数平滑预测算法预测将来一段时间内数据块的访问频率，在真实数据的实验结果上验证了预测方法的有效性。

关键词: 透明计算, 用户访问行为, 信息熵, 指数平滑, 行为预测