DocumentCode :
1869039
Title :
Evolutionary nonlinear data transformation for visualization and classification tasks
Author :
Zabkiewicz, Kamil
Author_Institution :
Inst. of Math. & Inf., Vilnius Univ., Vilnius, Lithuania
fYear :
2013
fDate :
8-11 Sept. 2013
Firstpage :
683
Lastpage :
685
Abstract :
In this paper we propose new approach in data set dimensionality reduction. We use classical principal component analysis transformation. Instead of rejecting features we generate new one by using nonlinear feature transformation. The values of transformation weights are changed evolutionary by using genetic algorithms. Results show better classification rates in smaller feature space. Visualization results also look better.
Keywords :
data reduction; data visualisation; feature extraction; genetic algorithms; pattern classification; principal component analysis; classical principal component analysis transformation; classification rate; classification task; data set dimensionality reduction; evolutionary nonlinear data transformation; feature generation; feature space; genetic algorithm; nonlinear feature transformation; transformation weight values; visualization task; Data visualization; Educational institutions; Electronic mail; Glass; Heart; Principal component analysis;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer Science and Information Systems (FedCSIS), 2013 Federated Conference on
Conference_Location :
Krako??w
Type :
conf
Filename :
6644080
Link To Document :
بازگشت