Research And Application Of Classification Algorithm Based On Decision Tree Rules

Data mining means the process of extracting cryptic and potential helpful information from a mass of Database. It is one kind of brand new data analysis technology and widely used in the fields of banking finance, insurance, government, education, transportation and national defense etc.Classification is one of important technique in Data mining. There are many methods for Data classification, in which the decision tree rules for the classification and prediction is a powerful tool. Generated decision tree rules, the rules usually order by two types: one is based rules ranking, called rule-based ranking; the other is based on class ranking, called class-based. Most of the algorithm is based on class ranking, such as the C4.5 rules algorithm, but for using class-based ranking, a poor quality of the rules may happen to predict the classes of higher rank, which may result in higher quality rules being ignored. The rule-based ranking can make up this shortcoming. This paper starts from the rule-based classification to make it more suitable for data mining application and to improve the classification accuracy. Main research works are as follows:First, this paper introduces data mining and describes the theoretical basis for classification, and some of the traditional classification algorithm.Second, a new classification algorithm based on rule-ranking is proposed, called CABRR algorithm. Three aspects will be considered when rank rules that are the length of the rules, the accuracy of the rules, and the coverage of the rules. Experiments results show that our method is higher effect than that of PART and C4.5 algorithm.Finally, the classification decision tree algorithm is applied to banks field for mining potential big customers, and the practicality of the algorithm has been analyzed.