摘要
Bagging是一种有代表性的机器学习的组合方法,它在改善弱分类器的稳定性和精度上有着重要的价值,已得到广泛的认可与应用.但它的重抽样技术使其在大数据挖掘中表现不够理想.mBagging是一种对Bagging加以改进的组合方法,克服Bagging一些不足,具有更高的统计功效、更低的假阳率和更快的运算速度.研究阐述mBagging的原理,从理论上探讨mBagging相较于Bagging性能更优的机理,并以皮尔逊相关系数作为基分类器,验证了mBagging的有效性.
Bagging is a representative of ensemble learning in machine learning.Its important value is to improve the robustness and accuracy of weak classifiers,and it has been widely accepted and applied.But the resampling technology of Bagging makes it be not suitable to big data mining.mBagging is an improved ensemble learning on Bagging,which overcomes parts of the shortcomings of Bagging and obtains higher statistical power,lower false positive rate and faster pace.This study elaborated the principle of mBagging,explored its mechanism,and employed Pearson correlation coefficient as the base classifier to verify its feasibility.
作者
刘汉明
刘赵发
郑金萍
胡声洲
LIU Hanming;LIU Zhaofa;ZHENG Jinping;HU Shenzhou(School of Mathematics and Computer Science,Gannan Normal University,Ganzhou 341000,China)
出处
《赣南师范大学学报》
2020年第3期33-35,共3页
Journal of Gannan Normal University
基金
国家自然科学基金项目(31660321)
江西省高校人文社会科学研究规划项目(YY17101)
江西省科技支撑计划项目(20171BBE50065)。
