Association Rules Algorithm And Measure Study

Data mining is the core of the KDD, data mining is found the interesting patterns from large amounts of data, that is to identify hidden, unknown, but it is very useful information from huge volumes of data in database. Association Rules is one of the important technology of data mining, this thesis is studied association rules, including the contact as follows:Firstly, this thesis is studied Apriori algorithm in-depth. Analyses algorithm thinking, for the algorithm flaw, this article is showed the improved algorithm separately, the example proves that the improved algorithm can effectively reduce the number of candidate item sets to improve the efficiency of implementation; effective in reducing the number of scan databases, reduces the I/O burden.Secondly, this thesis is studied FP-Growth algorithm thinking and FP-Tree storage structure in-depth. By analysing the algorithm, obtains the factors which affected the size of the FP-Tree, ensure to get the most compresed FP-Tree, thereby reduces I/O burden. The performance is compared between Apriori algorithm and FP-Growth algorithm through experiments.Thirdly, this thesis is studied metric method of association rules in-depth. From the two aspects of mathematics and measurement rules analyse the problems of interest measure, with the problems of interest measure, combines with the effectiveness of the principle of the rules, and proposes a measure improvement method. The results indicate that this improvement method can shield the invalid rule effectively.Finally, this thesis is showed an application example of the association rules, the algorithm is introduced to the bank customer information system, the experimental results show that the improvement measure can be effectively restrained the rules, and the improved method is effective.