Searching for optimal classifier using a combination of cluster ensemble and kernel method

This work introduces a supervised classification algorithm based on a combination of ensemble clustering and kernel method. The main idea of the algorithm lies behind the expectation that the ensemble clustering as a preliminary stage would restore more accurately metric relations between data objects under noise distortions and existence of complex data structures, eventually rising the overall classification quality. The algorithm consists in two major steps. On the first step, the averaged co-association matrix is calculated using cluster ensemble. It is proved that the matrix satisfies Mercer's condition, i.e., it defines symmetric non-negative definite kernel. On the next step, optimal classifier is found with the obtained kernel matrix as input. The classifier maximizes the width of hyperplane's separation margin in the space induced by the cluster ensemble kernel. Numerical experiments with artificial examples and real hyperspectral image have shown that the proposed algorithm possesses classification accuracy comparable with some state-of-the-art methods, and in many cases outperforms them, especially in noise conditions.