Mining frequent patterns directly from spatial datasets
Sha Li · Jisuanji yingyong yanjiu · 2013
To simplify the preprocessing procedure of spatial frequent pattern mining and enhance the efficiency,this paper proposed a mining algorithm called FISA.Using FISA frequent patterns(predicate sets and association rules) could be directly extracted from spatial datasets.Unlike transaction based mining algorithms,the complex and expert-dependent preprocessing procedure was not needed because FISA calculated the support of predicate sets using spatial intersect operation and area calculation instead of the record counting.Besides,vector or raster layers corresponding to predicate sets would be created during mining,which could be further used for visualization of mining results.Compare to other spatial analysis based mining algorithms,FISA supports both vector and raster layers,which were the majority format of spatial datasets.Also,it introduced a fast intersect method which could decrease the time complexity of support calculation into FISA to assure its scalability.Experimental results demonstrate that FISA is capable of mining frequent patterns from spatial datasets directly,correctly and efficiently.