基于大数据不平衡样本集的重采样方法及应用  

Resample Methods and Application Based on Imbalance Sample Set within Big Data

作　　者：汪海涛 余永奎 段春雨 WANG Hai-tao;YU Yong-kui;DUAN Chun-yu(Zhongshan Administration of Power Supply of Guangdong Power Grid,Zhongshan 528400)

机构地区：[1]广东电网有限责任公司中山供电局,中山528400

出　　处：《现代计算机》2018年第15期26-29,共4页Modern Computer

基　　金：广东电网有限责任公司科技项目(No.032000KK52160028)

摘　　要：讨论大数据不平衡样本集的5种重采样方法,对电力违章事故事件大数据不平衡样本集进行平衡处理,并讨论不平衡样本集训练机器学习算法的多个性能指标;以此为应用大数据不平衡样本集训练和比较机器学习算法,以及为选择性能较好的机器学习算法提供决策信息;所阐述的研究方法,对大数据分析及机器学习算法分类器模型的研究及应用具有普遍的参考价值和意义。Discusses five resample methods of imbalance sample set in big data,conducts equalization treatment on imbalance sample set in big data of electric power violation incidents.Also discusses several performance indexes of machine learning algorithm trained through imbalance sample set.Works for training application of imbalanced sample set in big data and comparison among machine learning algorithms,as well as provides decision-making information for selecting machine learning algorithm with better performance.The research method interpret.ed hereby is of general reference value and significance for the analysis of big data,and the research and application of machine learning al.gorithm classifier model.

关 键 词：大数据 重采样 不平衡 机器学习 分类器模型 

分 类 号：TM732[电气工程—电力系统及自动化] TP181[自动化与计算机技术—控制理论与控制工程]