基于Map-Reduce框架的C4.5分布式改进算法  

Animproved C4.5 distributed algorithm based on map-reduce framework

作　　者：廖绍雯[1] 贾聪 LIAO Shaowen;JIA Cong(Hexi College,College of Information Technology and media,Zhangye Gansu 734000,China;Inner Mongolia University,Computer Science Institute,Inner Mongolia 010021,China)

机构地区：[1]河西学院信息技术与传媒学院,甘肃张掖734000 [2]内蒙古大学计算机学院,内蒙古010021

出　　处：《自动化与仪器仪表》2020年第8期17-21,共5页Automation & Instrumentation

基　　金：内蒙古自治区高等学校科学研究项目(No.NJZY16383)。

摘　　要：为了利用决策树分类算法解决大规模数据集上的分类问题,提出了一种基于经典C4.5决策树算法的分布式算法,在Map-Reduce分布式计算系统框架下,利用分布式思想构造决策树节点,解决了数据量大和数据复杂度高导致分类时运行速度慢的问题。仿真结果表明,与经典C4.5算法相比,该分布式算法具有较高的分类准确率及较好的speedup、scaleup和sizeup性能。To take advantage of the decision tree classification algorithm to solve classification problems in large-scale da-ta set,a distributed algorithm is proposed based on classical C4. 5 decision tree algorithm. In the Map-Reduce under theframework of distributed computing systems distributed idea is used to construct the decision tree node,and the data volumeand high data complexity leading to slow classification of running speed is solved. The simulation results show that comparedwith the classical C4. 5 algorithm,the feasibility of the distributed algorithm has higher and better speedup,scaleup and sizeupperformance.

关 键 词：经典C4.5算法 大数据 分布式算法 MAP-REDUCE 准确率 

分 类 号：TP301􀆱[自动化与计算机技术—计算机系统结构]