Locality Preserving Projections Algorithm Based on Improved Iterative SelfOrganize Data Analysis

The new locally preserving projections algorithm is proposed in this paper which is based on Bayesian criteria and adapted improved iterative self-organize data analysis. The experiment shows that the new algorithm can put forward the optimum number of dimensions and be more available than principle component analysis. That is because it takes into account the relation the number of between dimensions and classification. The new algorithm not only preserves the structure of original data and eliminates the correlation and redundancy of high dimension vectors.