Boosting Principal Component Analysis by Genetic Algorithm

  • Divya Somvanshi Banaras Hindu University, Varanasi
  • R D S Yadava Banaras Hindu University, Varanasi
Keywords: Feature extraction, genetic algorithm, feature fusion, principal component analysis, pattern recognition


This paper presents a new method of feature extraction by combining principal component analysis and genetic algorithm. Use of multiple pre-processors in combination with principal component analysis generates alternate feature spaces for data representation. The present method works out the fusion of these multiple spaces to create higher dimensionality feature vectors. The fused feature vectors are given chromosome representation by taking feature components to be genes. Then these feature vectors are allowed to undergo genetic evolution individually. For genetic algorithm, initial population is created by calculating probability distance matrix, and by applying a probability distance metric such that all the genes which lie farther than a defined threshold are tripped to zero. The genetic evolution of fused feature vector brings out most significant feature components (genes) as survivours. A measure of significance is adapted on the basis of frequency of occurrence of the surviving genes in the current population. Finally, the feature vector is obtained by weighting the original feature components in proportion to their significance. The present algorithm is validated in combination with a neural network classifier based on error backpropagation algorithm, and by analysing a number of benchmark datasets available in the open sources.

Defence Science Journal, 2010, 60(4), pp.392-398, DOI:

Author Biography

Divya Somvanshi, Banaras Hindu University, Varanasi

Received her BSc from the Government Degree College, Hardoi, and MSc(Physics) from the DAV College, Kanpur, in 2005 and 2007, respectively. She is working as a Research fellow in a DRDO-ponsored research project on development of data fusion models and algorithms for electronic nose systems at the Department of Physics, Banaras Hindu University, Varanasi.

How to Cite
Somvanshi, D., & Yadava, R. D. S. (2010). Boosting Principal Component Analysis by Genetic Algorithm. Defence Science Journal, 60(4), 392-398.