几何平均参与评价划分属性的决策树  

Geometric average to participate in evaluation of attribute to build decision tree

作　　者：王卓[1] 聂斌[2] 罗计根 WANG Zhuo 1, NIE Bin 2, LUO Ji-gen 2(1. School of Software, Nanchang University, Nanchang 330047, China;2. School of Computer, Jiangxi University of Traditional Chinese Medicine, Nanchang 330004, Chin)

机构地区：[1]南昌大学软件学院,江西南昌330047 [2]江西中医药大学计算机学院,江西南昌330004

出　　处：《计算机工程与设计》2018年第7期1877-1882,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61562045);江西省教育科学"十二五"规划一般课题基金项目(15YB005);江西省卫生计生委中医药科研计划课题基金项目(2017A282)

摘　　要：针对信息增益偏向于多值属性,信息增益率倾向于少值属性的特点,研究几何平均参与评价划分属性的决策树。从候选划分属性中,筛选高于信息增益算术平均水平的属性;分别计算这些属性的信息增益与信息增益率的几何平均值,从中选择几何平均值最大的属性,建立分支决策;用递归方法建立决策树。对4份不同规模数据进行实验验证,验证结果表明,该决策树准确性较好,运行时间较低,可行有效。Focusing on the problems of information gain and information gain rate for the number of attribute values,an improved decision tree algorithm combining geometric average to participate in the evaluation of attribute was studied.The information gain of the candidate partitioning attribute was calculated to find out the attribute of the information gain higher than the average level.The geometric average of the information gain and information gain of these attributes were calculated respectively,the attribute with the largest geometric value was selected,and branch decision was established.Recursive method was used to establish decision tree.Through four experiments with different data scales,the results show that the proposed method has high accuracy and low running time,and has certain advantages.

关 键 词：信息增益 信息增益率 筛选 几何平均 决策树 中医药信息 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]