Research On Algorithms For Association Rules Based On Granular Computing And Complete Graph

As an emerging interdiscipline,the main task of Data Mining is to research and explore effective methods of information extraction and extract useful information hidden from huge databases.Based on the analysis of the traditional association rule algorithms,it is found that most of them generating frequent itemsets from candidate itemsets and having pattern matching through scanning the entire database more than one.In order to improve the efficiency of algorithms,this paper replaces the traditional pattern matching to Granular Computing(GRC) because of the characteristics of its low cost.At the same time,in order to avoid multi-pass scanning the entire database,the complete graph is used to divide mining region and mining frequent itemsets only in possible scope.The main research in this paper can be classed as follows:1.GRC_G algorithm(algorithm based on GRC and complete graph).At the base of learning from others,analysis the theory of GRC,and apply the idea to association rules mining.Put forward two new concepts:binary granule and complete combination of binary granule.And a method of reduce mining space based on complete graph is bring forward.At last,give the algorithm and the simulation tests show that this algorithm has better performances.2.T_GRC_G algorithm(algorithm for mining two-way association rules based on GRC_G).Analysis the vary characteristics of attributes in commonly actual database, we can see,some of which contains the relationship between attributes often follow certain rules of co-exist in pairs,but simply frequent itemsets mining can not find these effective rules.Therefore improve GRC_G algorithm and propose a mining algorithm for two-way association rules-T_GRC_G algorithm.Build the concepts of strong two-way association rules and strong-weak association rules.And in order to reduce the redundant rules generated,a method to delete redundant rules is put forward.The algorithm is given and simulation tests show that the algorithm can effectively reduce the redundant rules generated,and can find maybe more meaningful strong-weak two-way association rules.3.MD_GRC_Galgorithm(GRC_G algorithm in multi-dimensional space).GRC_G algorithm may be effective in many cases.However,in many situations,people are interested in knowledge in the multi-dimensional space.For this put forward this algorithm for mining multi-dimensional association rules based on GRC_G.It can mine multidimensional frequent item sets through projection and then selecte the association rules.Give.the algorithm,and the simulation results show that the algorithm can effectively discover multidimensional association rules and has higher time efficiency.4.In order to better prove that the improved algorithms proposed in this paper have effectiveness and value,at the same time of test by laboratory simulation,have an actual mining test in the database of traditional Chinese medicine(TCM) prescription.The results of test show that the improved algorithms are indeed able to effectively find interesting association rules in the practical application.At present,the field of TCM has not the perfect and mature precedent of research by introducing the theory of data mining in.