A Clustering Algorithm via Kernel Function and Locality Preserving Projections

This paper combines kernel function and Locality Preserving Projections (LPP) in a framework to improve the accuracy of clustering model. Specifically, we first use the kernel function to project each feature into high-dimensional kernel space so as to mine the nonlinear relationship of data. At the same time, the l 1 -norm sparse regularization term is used for feature selection. Besides, LPP saves the local structure of data. Finally, the optimal clustering result is obtained by solving the objective function. Experimental results on six benchmark data sets show that our proposed method is superior to the compared methods in term of clustering tasks.