Efficient Mining of High Utility Co-location Patterns Based on a Query Strategy.

A high utility co-location pattern (HUCP) is a set of spatial features, which is supported by groups of neighboring spatial instances, and the pattern utility ratios (PUR) of the spatial feature set are greater than a minimal utility threshold assigned by users, can reveal hidden relationships between spatial features in spatial datasets, is one of the most important branches of spatial data mining. The current algorithm for mining HUCPs adopts a level-wise search style. That is, it first generates candidates, then tests these candidates, and finally determines whether the candidates are HUCPs. It performs mining from the smallest size candidate and gradually expands until no more candidates are generated. However, in mining HUCPs, the UPR measurement scale does not hold the downward-closure property. If the level-wise search style is adopted, unnecessary candidates cannot be effectively pruned in advance, and the mining efficiency is extremely low, especially in large-scale and dense spatial datasets. To overcome this, this paper proposes a mining algorithm based on a query strategy. First, the neighboring spatial instances are obtained by enumerating maximal cliques, and then these maximal cliques are stored in a hash map structure. Neighboring spatial instances that support candidates can be quickly queried from the hash structure. Finally, the UPR of the candidate is calculated and a decision is made on it. A series of experiments are implemented on both synthetic and real datasets. Experimental results show that the proposed algorithm gives better mining performance than existing algorithms.