基于K-means聚类和遗传算法的少数类样本采样方法研究  

Research of Minority Kind of Sample Sampling Method Based on K-means Cluster and Genetic Algorithm

作　　者：杨永[1] 王莉利[1] 

机构地区：[1]大庆石油学院计算机与信息技术学院,大庆163318

出　　处：《科学技术与工程》2010年第10期2334-2338,共5页Science Technology and Engineering

基　　金：黑龙江省教育厅科学技术研究项目(11531013)资助

摘　　要：传统的分类器对不均衡数据集的分类严重倾向于多数类。为了有效地提高不均衡数据集中少数类的分类性能,针对此问题提出了一种基于K-means聚类和遗传算法的少数类样本采样方法。通过K-means算法将少数类样本聚类分组,在每个聚类内使用遗传算法获取新样本并进行有效性验证,最后通过使用KNN和SVM分类器,在仿真实验中证明了方法的有效性。The classification favors seriously to the most kinds when the traditional sorter is used to classify the imbalanced data set.In order to effectively enhance classified performance of the minority kind in the imbalanced data set,one kind minority kind of sample sampling method based on the K-means cluster and the genetic algorithm in view of this question is proposed.K-means algorithm to cluster and group the minority kind of sample is used,and in each cluster the genetic algorithm is used to gain the new sample and to carry on the valid confirmation.Finally,through using KNN and SVM sorter the method validity is proved in the simulation experiment.

关 键 词：K-MEANS算法 聚类 遗传算法 不均衡数据集 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]