Research On Semi-supervised Spectral Clustering Algorithm Based On The Optimal Projection

Research on the problem of high time and space complexity,obviously declining of clustering performance even disabled when spectral clustering solves the high dimensional and large quantity of data,semi-supervised spectral clustering based on the optimal projection was proposed.Analysis the supervisory information using the LDA criterion,then get the optimal projection vector space and project the original data to reduce dimensions,distinguish the importance of dimensions and improve clustering accuracy.At the same time,using the Nystr?m sampling algorithm to exchange similar matrix calculation into small amount of datas approach the whole samples,reduce the amount of calculation and improve the computational efficiency.Different kinds of datasets are selected from authority clustering and classification dataset UCI in this paper,the algorithm was compared with other algorithms,the results shown that the algorithm of this paper has higher clustering accuracy and computational efficiency,especially in the face of high dimension and large amount of data clustering problem.