基于云计算平台的代价敏感集成学习算法研究  

Study on the cost-sensitive ensemble learning algorithm based on the cloud computing platform

作　　者：张伶卫[1] 万文强[1] 

机构地区：[1]南京邮电大学计算机学院,江苏南京210003

出　　处：《山东大学学报（工学版）》2012年第4期19-23,28,共6页Journal of Shandong University（Engineering Science）

基　　金：国家重点基础研究发展计划(973计划)资助项目(2011CB302903);国家自然科学基金资助项目(61073114);南京邮电大学攀登计划资助项目(NY210010)

摘　　要：针对现实生活中大规模不平衡数据的分类问题,设计了一种基于云计算平台的代价敏感集成学习分类算法。Hadoop云计算平台对海量数据进行划分用于并行学习,同时结合代价敏感的思想对学习得到的基分类器进行加权集成,实现了云计算平台上的代价敏感集成学习分类模型。仿真实验表明该模型能够明显提高少数类的查全率,同时Hadoop的并行机制使得云平台坏境下的集成学习时间较集中式环境有大幅度的缩减,进一步提高了大规模不平衡数据分类问题的学习效率。With respect to the classification of large scale imbalanced data, a distributed cost-sensitive ensemble learning algorithm based on cloud computing platform was proposed. The large scale data was divided on Hadoop cloud compu- ting platform and was used in parallel learning. Based on the idea of cost-sensitive, a weighted ensemble classifier was achieved, and a distributed cost-sensitive ensemble learning model based on cloud computing platform was developed. Experiment results showed that the recall rate of the minority class was improved significantly and the computational time was shortened by the ensemble learning on cloud computing platform due to the Hadoop parallel mechanism. In ad- ditron, the classification efficiency of the large-scale imbalanced problem was largely improved.

关 键 词：代价敏感 集成学习 云计算平台 不平衡分类 分布式 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]