基于WEKA平台的决策树算法比较研究  

Comparative Study on Decision Tree Algorithm Based on WEKA Platform

作　　者：杨小军 钱鲁锋 别致 YANG Xiaojun;QIAN Lufeng;BIE Zhi(Logistics and Equipment Information Resources Teaching and Research Office,Joint Logistics College,National Defense University,Beijing 100858)

机构地区：[1]国防大学联合勤务学院后勤与装备信息资源教研室,北京100858

出　　处：《舰船电子工程》2018年第10期34-36,97,共4页Ship Electronic Engineering

摘　　要：决策树是数据挖掘领域广泛研究和应用的一种分类算法,具备计算量小、速度快、分类准确率高、分类规则易于理解等众多优点。论文选取了八个公开的UCI科研数据集,从分类准确率、建模速度、可解释性三个方面对经典的决策树算法C4.5、CART和NBTree进行比较,分析了三个算法各自的原理和优缺点,明确了各算法的适用情况。Decision tree is a kind of classification algorithm which is widely researched and applied in data mining. It hasmany advantages such as small computation,fast speed,high classification accuracy and easy to understand classification rules.This paper selects eight open UCI scientific data sets,and compares the classic decision tree algorithm C4.5,CART and NBTreefrom three aspects：classification accuracy,modeling speed,and interpretability. In this paper,the principles and advantages anddisadvantages of the three algorithms are analyzed,and the application of each algorithm is clarified.

关 键 词：决策树算法 C4.5 CART NBTree 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]