决策树ID3算法中引入简单工厂模式的设计研究  

Design Research of Decision Tree ID3 Algorithm Using Simple Factory Pattern

作　　者：孙道远[1] SUN Dao yuan(Department of Compmer Information Engineering, Anhui Vocational Technical College of Industry % Trade, Huainan Anhui 232001, China)

机构地区：[1]安徽工贸职业技术学院计算机信息工程系,安徽淮南232001

出　　处：《德州学院学报》2018年第2期61-64,共4页Journal of Dezhou University

基　　金：安徽省高校自然科学研究项目(KJ2017A653);安徽省高等学校省级质量工程项目(2015sjjd048)

摘　　要：根据经典决策树ID3算法,通过采集确定的训练样本集构造决策树模型,在判断决策树节点属性过程中针对期望值及熵的计算引入模式设计GoF中简单工厂模式设计一个工厂类,在工厂类静态方法计算出该节点的样本特征变量对象与类别特征变量对象,把这两种特征变量对象返回的期望值传递给ID3算法中递归过程,从而构造出决策树模型,相比较在ID3算法中把期望值的计算集成在递归中,这种松耦合的封装分离方式可方便各种评测环境下的节点扩充,增强了计算效率和代码的维护性,以及提高算法在其他应用中的移植性.According to the classical decision tree ID3 algorithm, the decision tree model is constructed from ＂the set of ＂training samples. This is related to ＂the calculation of ＂the expected value of ＂the expected value of ＂the characteristic variable and ＂the expected value of ＂the variable in ＂the sample set. In ＂the tradition al algorithm, And ＂the calculation of entropy is often integrated in ＂the recursive process of ＂the ＂training sam pie set, so ＂the algorithm to achieve ＂the code to increase ＂the portability and maintenance of ＂the difficulty. In ＂the decision ＂tree node attribute process for ＂the expected value and ＂the calculation of ＂the entropy into ＂the model design GoF simple factory model design a plant class , in ＂the factory class static method to calculate ＂the node of ＂the sample characteristics of ＂the variable object and category characteristics of variable objects, The expected value of ＂the return of ＂these ＂two characteristic variables is passed to ＂the recursion process in ID3 algorithm to construct ＂the decision ＂tree model. Compared with ＂the ID3 algorithm, ＂the calculation of ＂the expected value is recursively. This loose coupling method can be convenient Node expansion in a varie ty of evaluation environments enhances computational efficiency and code maintenance, as well as improves ＂the portability of algorithms in other applications.

关 键 词：决策树 简单工厂 ID3 期望值 熵 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]