Most statistic machine learning and pattern recognition problems begin with high-dimensional data processing. ”Looking” at the high-dimensional dataset plays an important role in data analysis, which can give analysts a direct impression of the structure of the dataset. Dimension reduction methods for visualizing data aim at creating a low-dimensional data representation of the high-dimensional dataset, which can be easily plotted or viewed by human beings. These low-dimensional representations require to exhibit the intrinsic structure of the high-dimensional dataset. This work proposes a data visualization algorithm which can both keep local and global structure of the high-dimensional dataset. We present a way to generate anchor points of the dataset. With these embedded anchor points, we proposed the ADR method, then combine theADR method with the LOE method to construct a bi-objective model. The embedding result can both preserve the global and the local information.