摘要
税收信用分类管理在税务系统中起着重要作用,应用分类算法解决税收信用等级手工评定问题是当前税务系统的难题之一。决策树算法是分类算法中一类重要算法,其中以C4.5算法最为经典,但该算法在连续属性离散化方面花费时间成本较多。该文在C4.5连续属性离散化算法基础上引入基于经验值的窗口分割技术,在保证生成决策树准确率的前提下,有效的提高了算法运行效率。应用改进算法构造税收信用等级判定决策树,并根据构造的决策树实现对纳税人税收信用等级的自动判决。
Taxation credit classification management plays an import role in tax system, and apply classify algorithm to solve the problem of taxation credit classification by manual operation is one of the difficult points for today's taxation system . The Decision Tree Algorithm, especially C4.5, is an important kind of classification algorithm, However, C4.5 doesn't do a very good job in continuous attributes discretization. This treatise introduced window split technology based on empirical value into C4.5 continuous attributes discretization algorithm, effectively improved the effectiveness of algorithm and guaranteed the accuracy of built decision trees at the same time. It also built a taxation credit grade judgment decision tree with improved algorithm and automatic judging the taxation credit grade by the generated decision tree.
出处
《微计算机信息》
2009年第15期264-266,共3页
Control & Automation
基金
基金申请人:胡小建教授
项目名称:基于网格的开放式决策支持方法与决策支持系统的研究
颁发部门:安徽省自然科学基金
安徽省自然科学基金委
合肥工业大学(070416241)
关键词
决策树
C4.5算法
税收信用分类
经验值窗口分割
Decision Tree
C4.5 algorithm
taxation credit classification
empirical value window split