Abstract
Data stream management is a hot topic of recent interests. Mining frequent itemsets has been widely studied in this field. However, most of the past works focus on mining frequent itemsets over a single data stream. Unfortunately, the research can be impractical in real applications, such as network management, market analysis, etc. In the last few years, several researchers have noticed this problem and proposed several approaches to address it. Mining frequent itemsets over multiple data streams is one of the major objectives of our work. Furthermore, we consider the request for frequent itemsets over multiple data streams as a special type of queries. We propose a framework for processing multiple queries over multiple data streams. The major difficulty of the work is to handle a large amount of data from several data streams to efficiently produce the mining results. Thus, an efficient processing plan is essential. In this paper, we introduce the concept of the Plan Graph for enumerating all possible processing plans. By using the Plan Graph, we transform the problem of finding the optimal processing plan into the problem of finding the minimum-cost tree on a graph. Although the Plan Graph enumerates successfully all possible processing plans, finding the optimal processing plan can be time-consuming. Therefore, we propose a heuristic approach to efficiently find a plan close to the optimal one. Finally, we verify the efficiency of our approach by the experiments.