Mining maximal frequent intervals
Jun-Lin Lin · 2003
Many real world data are associated with intervals of time or distance. Mining frequent intervals from such data allows the users to group transactions with similar behavior together. Previous work only focuses on the problem of mining frequent intervals in a discrete domain. This paper first proposes the notion of maximal frequent intervals, which are superior to frequent intervals in many perspectives, and then provides a method for mining maximal frequent intervals in either a discrete domain or a continuous domain. Experimental results indicate that our method outperforms previous method for mining frequent intervals, and improves the runtime by two orders of magnitude for databases with a large number of distinct endpoints.