The Application's Research Of Data Mining In Stockjobber's CRM

Data Mining is a new technology in the rapid development information technology. It is used to extracting the useful information and knowledge from application data. Classification is a very important branch of data mining field. And it has been discussed by many researchers. Decision tree is the most important field in classification. Although various classify algorithms suitable for decision tree have been put forward, there exist some problems with them, such as computing efficiency, stability, scalability and so on. Therefore classify algorithm needs deeper research to meet engineering demanding. And consequently, a new Rough Decision tree be proposed in this thesis and their high performance compared to C4.5 algorithm is demonstrated by some examples. In addition, designing a system model of stockjobber' s CRM according to the characteristic of securities industry in china. The major achievements of this thesis are as follows.Firstly, to expatiate the classify processing, and described some main technologies used in classify detail. It also summarize the evaluation standard of classify.Secondly, study of ID3 algorithm and C4.5 algorithm. Some test and conclusion has been drawn by an actual example. Besides, compare the characteristics of above algorithms.Thirdly, give some examples according to the feature of customer data, analyse it with decision tree,Naivebayes and RBFNetwork algorithm. It also analyse the classification performance by the four algorithms. And summarize the reason why choose decision tree and rough set to combine.Fourthly, study the theory of attribute importance in Rough set. Combining the C4.5 algorithm, which take information gain ratio as attribute choice criterion. In this thesis try to adopt attribute importance method to alternate information gain ratio as attribute choice criterion. The new approach that based on Rough set and Decision tree is proposed.Finally, in this thesis, it describes in detail several definition and characteristics of CRM. According to the characteristics of securities industry in china designed a system model of stockjobber's CRM. Then a star-shaped data warehouse model has been built. The new rough decision tree method is applied to division clients. The experimentations manifest that applying the new method to classification such as division clients in securities industry has a good result. In addition, puts forword some suggestions about the application of data mining in stockjobber' s CRM.