Research And Application Of Data Mining Techniques Based On CRM

The requirement of application for mining information from the amount of data, and the influence which the information has on the strategic decisions of enterprises, make data mining significative either the theorical research or the practice. This dissertation studies data mining technique and its application in Customer Relationship Management (CRM).Mining for association rules is an important embranchment of data mining. It has a lot of advantages such as understandability, intelligibility, sententiousness, and broad application. The main subject is to find interesting association or correlation relationships among a large set of data items. Finding all frequent itemsets is the first step of association rule mining. The main method of realization usually is algorithm like Apriori to find frequent itemsets. But efficiency of the Apriori algorithm needs to be improved.The research about mining maximal frequent itemsets is very important because maximal frequent itemsets contain all frequent itemsets and some application only need to find maximal frequent itemsets. An algorithm for mining maximal frequent itemsets, called MFIAVTL (Maximal Frequent Itemsets AlgorithmVertical Tid-List), is presented in this dissertation aiming at the characteristic of CRM data. MFIA_VTL employs a Vertical Tid-List database layout scheme. Along with depth first search strategy, it uses a partitions method based on prefix to divide up search space into lesser subspace. A subspace is a potential maximal frequent itemset. Finding maximum frequent itemsets can be executed in a superset that contains maximum frequent itemsets as far as possiable small. It will reduce I/O spending obviously. The vertical Tid-List database layout scheme makes it avaliable to count the support of itemsets simply by set intersection operations instead of scanning database repeatedly. The experiment shows that the MFIAVTL algorithm is stable, scalable and effective.