Spectroscopy and Spectral Analysis, Volume. 36, Issue 4, 1245(2016)
Study on Stellar Spectral Outliers Mining Based on Fuzzy Large Margin and Minimum Ball Classification Model
It’s one of the main goals in universe exploration to find unknown and special celestial bodies. The spectra outlier data is analyzed based on the traditional classification approaches, which is a general method of special celestial body exploration. But it’s depressed that many traditional classification approaches are insensitive to the outlier data, which even influence the classification efficiencies, therefore, these methods can’t accomplish the task of special celestial body exploration. In view of this, Fuzzy Large Margin and Minimum Ball Classification Model (FLM-MBC) is proposed in this paper. In FLM-MBC, part of general data and outlier data are trained to construct the minimum ball model and the fuzzy technique is introduced to reduce the noise influence to classification. Comparative experiments with C-SVM, KNN, and SVDD on the SDSS spectral datasets verify the effectiveness of the proposed FLM-MBC.
Get Citation
Copy Citation Text
LIU Zhong-bao, ZHAO Wen-juan. Study on Stellar Spectral Outliers Mining Based on Fuzzy Large Margin and Minimum Ball Classification Model[J]. Spectroscopy and Spectral Analysis, 2016, 36(4): 1245
Received: Feb. 16, 2015
Accepted: --
Published Online: Dec. 20, 2016
The Author Email: Zhong-bao LIU (liu_zhongbao@hotmail.com)