Research On High Utility Co-location Patterns Mining

Spatiotemporal co-location pattern(hereinafter referred to as co-location)mining is a very popular field in spatial data mining,which means that multiple object instances are often in the neighboring state at same time.Most of the existing algorithms in this field use the method of item set mining of transaction database to mine the co-location mode of spatial entities.In the initial frequent co-location pattern mining,only the frequency of the occurrence of each spatial isomorphic pattern in the whole spatial range was considered,and the internal output of each spatial entity was not considered.Thus,many co-location patterns with high output and low occurrence rate could not be excavated.The introduction of the high utility co-location pattern mining(High Utility Co-location Pattern Mining,HUCPM)effectively solves the shortcoming of focusing only on the frequency of occurrence and not on the profit output,and it can dig out a more valuable co-location model to guide scientific decisions and extract the implied predictive information.In terms of innovation,this paper first uses the degree method to find maximum groups,and divides the space points in the whole region into several disjoint maximal groups.At the same time,KD-tree structure is used to improve the method of finding the maximal clique,so that the efficiency is improved significantly.The point set divided into several disjoint maximal groups can be regarded as a disjoint transaction set.The utility of the extended mode of transaction--that is the total utility of all the instance points in a single maximal clique can be calculated and used as pruning,so that mining can be conducted in the same way as mining the high utility itemset.Mining the co-location pattern efficiently is similar to mining the frequent item set in the transaction database in that the downward closing property of frequency is used as a pruning strategy.However,the efficient use of frequent item set does not have the good property of the frequent item set,that is,inverse monotonicity;This article finally adopt the method of mining efficiency with set based on subsume index(,IHUI-Mine),to the great mass of transaction collection divide before out to represent business set with bitmap matrix,and then with the help of every great group total utility monotonicity,avoid to use more based on the table before instance method,not only improved the pruning of without the monotonicity problem,also can optimize the problem of repeated traversal of the original database,in the mining efficiency at the same time,memory usage has a better performance and scalability.In the end,the paper use graph display to visually display the excavated efficient co-location mode,which can more vividly reflect the utility obtained by the neighboring relations among different spatial transactions.