Research Of Data Mining Algorithm Based On Association Rules

Data Mining or Knowledge Discoverying is a technology used in data analysis, data understanding,discoverying knowledge contained in the data.It has been researched widely in recent years .Association rules is an important question in data mining,which must experience conception'bringing forward,conception'acceptance,extensive research and exploration, gradual application and substantive application.At present,the basic conception and research measures of association rules are becoming much clearer and developing towards deeper direction.From current situation,the most of scholars think that the research of association rules is still in a phase of extensive research and exploration,needs more innovation in basical theories, application models and mining algorithm.Meanwhile,association rules needs to be improved in mining efficiency,usability,accuracy.Hence,researchers need explore new mining theories and models,improve traditional algorithm, study new efficient algorithm.With regard to the current situation as well as development of data mining and association rules ,I select this topic to carry out correlative work.In association rules theories,basic conception,typical algorithm as well as some new research of association rules were sorted, concluded and summarized comprehensively,Some applied conditions and differences among different algorithm were compared objectively in association rules algorithm,In order to meet the high requirements for the memory and CPU in the process of mining large-scale data,bit string organization was proposed to organize data,which can improve storage efficiency in entire data mining.Two data structures algorithm,bit string trees and positive-negative association rules of frequent pattern trees were proposed in association rules.Bit string trees will compress large data into a small data structure to count support and confidence of 2 items;Then ,frequent items tree was built from top to bottom using support and confidence of 2 items,all the association rules will be mined.In order to avoid limitation of only mining positive association rules,positive-negative association rules based on frequent pattern tree was also proposed ,which will deal with positive items and hidden negative items in the database, breaking single pattern of mining positive association rules firstly, positive association rules secondly.The two data strctures algorithm will mine all the positive and negative association rules,scaning database only once without producing candidate items.The test indicated that algorithm had better capability in mining efficency,usability and expansibility.The algorithm will have applied value,on the one hand,it will provide scientific researchers studying association rules with thoughts for reference,on the other hand,the applied value will be enhanced if mingled in data mining tools.