Nonlinear dimensionality reduction method is more sensitive to the noise in the high-dimensional data, resulting in relatively poor final results of classification. In order to make up for its shortcomings, this paper proposes a method in the premise of using the maximum likelihood estimation method to estimate the intrinsic dimension of the sample data, which combines the isomapetric mapping with the principal component analysis. On the one hand, the method enables the original data to maintain its geometry in the high dimensional space, on the other hand, the method can eliminate the influence of noise on the dimensionality reduction results, eventually making the low-dimensional data as much as possible to maintain inherent characteristics of the original sample data sets. Experimental demonstrations show that the results of combination method is better than separate isometric mapping and separate principal component analysis.