Mining confident co-location rules without a support threshold

Mining co-location patterns from spatial databases may reveal types of spatial features likely located as neighbors in space. In this paper, we address the problem of mining confident co-location rules without a support threshold. First, we propose a novel measure called the maximal participation index. We show that every confident co-location rule corresponds to a co-location pattern with a high maximal participation index value. Second, we show that the maximal participation index is non-monotonic, and thus the conventional Apriori-like pruning does not work directly. We identify an interesting weak monotonic property for the index and develop efficient algorithms to mine confident co-location rules. An extensive performance study shows that our method is both effective and efficient for large spatial databases.