Abstract: By considering the correlation of the data and the using the clustering technique of pattern recognition, the relationship among various corrosive factors of atmospheric environment has been analyzed. The results of the clustering show that the 21 atmospheric corrosive factors are clustered into five distinct clusters under the correlation coefficient value above 0.54.On the basis of the clustering analysis, the abilities of corrosive factors to distinguish atmospheric corrosivity are assessed, and the main factors are selected. In the generalized space of the main factors, the disposition of the nearest neighbors for the experimental samples are analyzed and the way to predict the atmospheric corrosivity for unknown samples is studied.