PARALLEL CLASSIFIERS ENSEMBLE WITH HIERARCHICAL MACHINE LEARNING FOR IMBALANCED CLASSES

摘要：

Imbalanced distributions and mis-classified costs of two classes made conventional classification methods suffered. This paper proposed a new fast parallel classification method for imbalanced classes. Considering imbalanced distributions, the approach adopted a fast simple classifier with less features input working parallel with a complicated one. Most samples would be correctly recognized by the first classifier, and the second relatively slower classifier could be ended. The second one was only trained and worked for less difficult samples. Experimental results in machine vision quality inspection showed that the approach could effectively improve classification speed and decrease total risk for imbalanced classes classification.

关键词： Pattern recognition Imbalanced classes Hierarchical machine learning Parallel processing ROC

作者: YUN ZHANG BING LUO

作者单位: Faculty of Automation, Guangdong University of Technology, Guangzhou 510006, China

会议类型: 国际会议

会议名称: 2008 International Conference on Machine Learning and Cybernetics(2008机器学习与控制论国际会议)

会议地点: 昆明

会议语种:英文

页码: 94-99

在线出版日期: 2008-07-12（万方平台首次上网日期，不代表论文的发表时间）