Data mining proxy: Serving large number of users for efficient frequent itemset mining

Data mining has attracted a lot of research efforts during the past decade. However, little work has been reported on supporting a large number of users with similar data mining tasks. In this paper, we present a data mining proxy approach that provides basic services so that the overall performance of the system can be maximized for frequent itemset mining. Our data mining proxy is designed to fast respond a user's request by constructing the required tree using both trees in the proxy that are previously built for other users' requests and trees stored on disk that are pre-computed from transactional databases. We define a set of basic operations on pattern trees including tree projection and tree merge. Our performance study indicated that the data mining proxy significantly reduces the I/O cost and CPU cost to construct trees. The frequent pattern mining costs with the trees constructed can be minimized