Supervised Classification Box Algorithm Based on Graph Partitioning

In this paper we introduce the supervised classification algorithm called Box algorithm based on feature space partitioning. The construction of Box algorithm is closely linked to the solution of computational geometry problem involving heuristic maximal clique cover problem satisfying the k-nearest neighbor rule. We first apply a heuristic algorithm to partition a graph into a minimal number of maximal cliques and subsequently the cliques are merged by means of the k-nearest neighbor rule. The main advantage of the new approach is decomposition of the l-class problem (\(l > 2\)) into l single-class optimization problems. The performance of the Box algorithm is demonstrated to be significantly better than SVM in computer experiments involving real Monk’s dataset from UCI depository and simulated normal data.

[1]  Ventzeslav Valev,et al.  Machine learning of syndromes for different types of features , 2011, 2011 International Conference on High Performance Computing & Simulation.

[2]  K. VijayaKumar,et al.  A feature partitioning approach to subspace classification , 2007, TENCON 2007 - 2007 IEEE Region 10 Conference.

[3]  Kien A. Hua,et al.  SubSpace Projection: A unified framework for a class of partition-based dimension reduction techniques , 2009, Inf. Sci..

[4]  Sameer A. Nene,et al.  A simple algorithm for nearest neighbor search in high dimensions , 1997 .

[5]  H. Altay Güvenir,et al.  Diagnosis of gastric carcinoma by classification on feature projections , 2004, Artif. Intell. Medicine.

[6]  Ventzeslav Valev,et al.  Supervised pattern recognition by parallel feature partitioning , 2004, Pattern Recognit..

[7]  Bahram Alidaee,et al.  Clique Partitioning for Clustering: A Comparison with K-Means and Latent Class Analysis , 2007, Commun. Stat. Simul. Comput..

[8]  Jon Louis Bentley,et al.  Multidimensional binary search trees used for associative searching , 1975, CACM.

[9]  Atul Negi,et al.  A Novel Approach to Eigenpalm Features Using Feature-Partitioning Framework , 2007, MVA.

[10]  Adam Krzyzak,et al.  A new geometrical approach for solving the supervised pattern recognition problem , 2016, 2016 23rd International Conference on Pattern Recognition (ICPR).

[11]  H. A. Guvenir,et al.  Classification by Feature Partitioning , 1996, Machine Learning.

[12]  Antonio Iovanella,et al.  Some Spin Glass Ideas Applied to the Clique Problem , 2006, cond-mat/0605336.

[13]  Atul Negi,et al.  SubXPCA and a generalized feature partitioning approach to principal component analysis , 2008, Pattern Recognit..

[14]  Nicola Yanev,et al.  A combinatorial approach to the classification problem , 1999, Eur. J. Oper. Res..

[15]  Atul Negi,et al.  Novel approaches to principal component analysis of image data based on feature partitioning framework , 2008, Pattern Recognit. Lett..

[16]  Zhiqiang Zhang,et al.  A Flexible Annealing Chaotic Neural Network to Maximum Clique Problem , 2007, Int. J. Neural Syst..

[17]  Ventzeslav Valev,et al.  Classification using graph partitioning , 2012, Proceedings of the 21st International Conference on Pattern Recognition (ICPR2012).

[18]  M. Pelillo,et al.  Payoff-Monotonic Game Dynamics and the Maximum Clique Problem , 2006 .

[19]  Ventzeslav Valev,et al.  Supervised Pattern Recognition with Heterogeneous Features , 2004, CIARP.

[20]  Rong Long Wang,et al.  An Efficient Approximation Algorithm for Finding a Maximum Clique Using Hopfield Network Learning , 2003, Neural Computation.

[21]  Ventzeslav Valev,et al.  From binary features to Non-Reducible Descriptors in supervised pattern recognition problems , 2014, Pattern Recognit. Lett..