| Association reflects the interdependent or associated knowledge between an eventand other events. If there aresome rules between two or much variables, it is named as association. Association rule is one of the main data mining means. Through searching all things of system, it finds out some high conditional probability modes between the data objects. Those modes are considered to be the association rules between the data objects. Association rules mining is finished by two steps. One is to find all frequent itemsets. The other is to produce strong association rules by frequent itemsets. The first step is the key step. It includes twocommon means, which are Apriori algorithm and FP-tree algorithm.It is the core and base for modern corporate management to make policy scientifically. Traditional data analysis of market research is to process data after survey asper individual item through statistics. While if want to explode the relation between things that is existent objectively but unknown, it need to find out the associated net hidden in the database. Association rules mining can find out the interesting association or correlation between itemsets from a lot of dates. Thereby, it is widely used in sorts of decision supporting system, expert system, intelligent information system and so on. The paper tries to use the skill of associated rule in the investigation of telecommunication Wide-Brand market, and founds a data-mining subsystem basing on Wide-Brand service, looked forward to found out the using rule of Wide-Brand user and provide information needed for policy-makers to make policy, and provide exact database for scientific policy-making by knowing the managing situation.According to the needs of telecommunication and current database, the paper designs associated rule mining system with three function modules, which are customer consumptive behavior analysis, complaint and consultation analysis, and leaving situation analysis. Then, the author detailedly introduces the whole process of setting up the system in SOL Server 2000, including mining assignment analyzing, system design, data pretreatment and algorithm realization and etc. Firstly, mining assignment analyzing make it certain that the functions that system needed to finish and the database needed to realize those functions. Secondly, the system design includes the hardware environment selection, algorithm selection, system model design, and database model design. Thirdly, data pretreatment mainly includes the selection, integration, clearing, and transform of dates. Fourthly, algorithm realization is to use the idea of Apriori calculation, to realize discovering of frequent item gather and the birth of associated rules through the process of PL/SQL storage. Finally, using the value evaluation method of association rules, the paper discusses the mining results to finds the shortage of the basic method and suggests improved models and methods of system implemental influence from, includes adding more interests measurement, increasing optional parameter setting, anglicizing and ascertaining the improved result and so on. Ultimately lead the system to get a better practical effect. |
