Research on improved normalized cut spectral clustering algorithm

For the character of traditional clustering algorithms can only deal with low-dimensional linear data, not sensitive to high-dimensional nonlinear data sets, this paper proposes to use normalized cut (N-cut) spectral clustering algorithm to process high-dimensional data, and combines with sparse subspace clustering algorithm to generate similarity matrix, the method overcomes the insensitivity of N-cut spectral clustering to nonlinear discrete data, the clustering results of experiment data sets demonstrates the feasibility and robustness of this algorithm.