摘要
不平衡数据的分类是机器学习的热点问题.传统的分类方法在分类时会倾向于多数类而使得分类精度不高.对不平衡数据集的分类,提出一种基于FCM结合KFDA方法,首先采用FCM算法对样本数据进行聚类,将数据聚类后的样本数据映射到特征空间里,再采用KFDA算法对数据进行分类,可以克服不平衡数据对分类性能的影响.对UCI数据集进行仿真实验,结果表明FCM-KFDA算法可以有效地提高数据识别率.
The classification of imbalanced data is a hot problem in machine learning,which can tend to the majority class sample and the classification accuracy is not high,when it is classified by traditional methods.A new method based on FCM and KFDA is proposed for imbalanced data sets.Firstly,the FCM algorithm is used to cluster sample data,the sample data clustered are mapped to the feature space,and then KFDA algorithm is used to classify the data.so it can overcome the influence of imbalanced data classification performance.Simulation experiments on UCI data sets were done,the results showed that the FCM-KFDA algorithm can effectively improve the recognition rate of the imbalancede data sets.
出处
《华中师范大学学报(自然科学版)》
CAS
北大核心
2013年第6期776-780,共5页
Journal of Central China Normal University:Natural Sciences
基金
江苏省科技厅项目(BN2011056)