文献类型：会议论文

英文题名：Research of Medical High-dimensional Imbalanced Data Classification-Ensemble Feature Selection Algorithm with Random Forest

作者：Zhu, Min Su, Bo Ning, Gangmin

第一作者：Zhu, Min

通信作者：Ning, GM[1]

机构：[1]Zhejiang Univ, Hangzhou 310058, Zhejiang, Peoples R China;[2]Guizhou Univ, Guizhou Key Lab Agr Bioengn, Guiyang 550025, Guizhou, Peoples R China;[3]Guizhou Inst Technol, Guiyang 550003, Guizhou, Peoples R China

第一机构：Zhejiang Univ, Hangzhou 310058, Zhejiang, Peoples R China

通信机构：corresponding author), Zhejiang Univ, Hangzhou 310058, Zhejiang, Peoples R China.

会议论文集：International Conference on Smart Grid and Electrical Automation (ICSGEA)

会议日期：MAY 27-28, 2017

会议地点：Changsha, PEOPLES R CHINA

语种：英文

外文关键词：Medical; High-dimensional Imbalanced Data; Classification; Feature Selection; Random Forest

年份：2017

摘要：The purpose of this research is to effectively classify medical data, to provide accurate data foundation for clinical diagnosis and pathology, thereby improving the prediction and discrimination of clinical diagnosis. Aiming at the problem of poor classification accuracy, caused by high-dimensional characteristics and interclass imbalance factors in clinical medical data, this paper explore a medical high-dimensional imbalanced data classification method based on random forest ensemble feature selection algorithm. A high-dimensional feature space distribution model of medical high-dimensional imbalanced data is constructed by using phase space reconstruction method. The dimension reduction is carried by K-L feature compression method, to achieve the ensemble feature optimization of medical high-dimensional imbalanced data. The selected characteristic quantity is feature-classified by random forest classification method, and the medical data with different attribute characteristics are output. The simulation results show that the proposed method can be applied to medical high-dimensional imbalanced data feature selection and classification, which has good classification accuracy, low error rate, strong anti-interclass interference capacity and good clinical application value.