SIMCAにおける説明変数選択の影響

貝原 巳樹雄; 稲葉 光一

doi:10.11545/ciqs.2005.0.J08.0

28th Symposium on Chemical Information and Computer Sciences, Osaka

DOI https://doi.org/10.11545/ciqs.2005.0.J08.0

Conference information

Host: Division of Chemical Information and Computer Science, The Chemical Society of Japan

Co-host: The Pharmaceutical Society of Japan, Japan Society for Bioscience, Biotechnology, and Agrochemistry, The Japan Society for Analytical Chemistry, Society of Computer Chemistry, Japan, Graduate School of Pharmaceutical Sciences, Osaka University, Japanese Society for Information and Systems in Education (Approaval)

Oral Session

Influence of variable selection in case of SIMCA (Soft Independent Modeling of Class Analogy)

*Mikio Kaihara, Khoichi Inaba

Author information

Keywords: simca, variable selection, NIR spectra, classification

CONFERENCE PROCEEDINGS FREE ACCESS

Pages J08

Details

Abstract

SIMCA (Soft Independent Modeling of Class Analogy) has several good points, for example, the capability of 2Dimensional mappings of multi dimensional samples or the F-tests for classifications. Therefore, if we could select important variables and realize better and clear classification, such sample selection method should be significantly effective and important for the SIMCA. However, we found that just using variables with high modeling powers did not work well. The more, the number of optimal factors should be, the more, such tendency increases. Therefore, decreasing the number of variables for the SIMCA analysis too much means that the misclassification rate should be increasing. The analogy is one of the most important points for classifications, but it could have the limit. Finally, we tried the other classification methods and compared them, each other.

Corresponding author

Register with J-STAGE for free!