An efficient cube structure for stream data analysis

Data streams are very important for many applications in different areas. Compared to the traditional data in RDBMS, data streams have different characteristics such as unlimited data size, high data speed, and flexible time intervals. In this paper, we introduce a stream cube structure to organize the aggregations of stream data. The main idea is building cuboids on exponential time frames, and using base cuboids to approximate values in any time intervals. We implement stream cube models using Hyper-Tree Forest data structure. Theory analysis and experiments demonstrate that stream cube structure and its implementation are effective and efficient to answer ad hoc queries.