Abstract: A spatial co-location pattern is a group of spatial objects whose instances are frequently located in the same region．There are lots of jobs and achievements of co-location patterns mining algorithms for certain and uncertain data,but less maximal co-location patterns mining algorithms,especially spatial maximal co-location patterns for fuzzy objects．Mevent-tree algorithm was proposed in this paper for mining maximal co-location patterns for fuzzy objects.Firstly,it builds a event tree which can get candidate patterns for each object,build the HUT tree of candidate patterns,and then depth-first searches maximal co-location patterns begining with maximal-size candidate patterns and ending to size-2candidate patterns in HUT tree,as well pruning co-location candidate patterns after geting maximal co-location patterns．Then we put forward two improved strategies,including the pruning fuzzy objects during preprocessing and the pruning co-location candidates before creating HUT trees．Finally,extensive experiments show the effectiveness and efficiency of Mevent-tree algorithm and its improved methods.

Key words: Fuzzy objects,Maximal co-location pattern mining,Fuzzy participation ratio