Clustering based polyhedral conic functions algorithm in classification

ÖZTÜRK G., Ciftci M. T.

Journal of Industrial and Management Optimization, vol.11, no.3, pp.921-932, 2014 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 11 Issue: 3
  • Publication Date: 2014
  • Doi Number: 10.3934/jimo.2015.11.921
  • Journal Name: Journal of Industrial and Management Optimization
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.921-932
  • Keywords: Classification, polyhedral conic cunctions, K-means, linear programming, computational learning theory, RADIAL EPIDERIVATIVES, SEPARATION
  • Anadolu University Affiliated: Yes


In this study, a new algorithm based on polyhedral conic functions (PCFs) is developed to solve multi-class supervised data classification problems. The k PCFs are constructed for each class in order to separate it from the rest of the data set. The k-means algorithm is applied to find vertices of PCFs and then a linear programming model is solved to calculate the parameters of each PCF. The separating functions for each class are obtained as a pointwise minimum of the PCFs. A class label is assigned to the test point according to its minimum value over all separating functions. In order to demonstrate the performance of the proposed algorithm, it is applied to solve classification problems in publicly available data sets. The comparative results with some mainstream classifiers are presented.