The object of this study is to increase the accuracy of the result of principal component analysis (PCA). PCA is a well known method to capture smaller uncorrelated dimensions, which are the principal components, from correlated observational high dimensions. The smaller dimensional space is obtained as the most explainable hyper plane space by orthogonal projection of data in observational high dimension space. However, since the explanatory power is evaluated by the relatively small distances from the objects to the hyperplane spanned by the vectors of the principal components and only the non-expansive property is satisfied for the fixed two objects between the distances in the obtained space and in the original observational space, it may happen that there is a significantly larger difference between the two distances. In order to combat this attitude, we propose a principal component analysis considered dissimilarity structure of objects in high dimensional space by adopting a fuzzy clustering method.