Abstract: Recently, industrial cluster has been a hot focus in economics, management and geography. Regional and industrial policies are also oriented towards the promotion of industrial clusters. The identification of local clusters, however, is still poorly established. Qualitative methods of identifying industrial clusters, such as Industry Perception Method, rely heavily on the experts' subjective judgments and lack strict rules to make decisions, therefore confine to a number of limitation. For example, the regionally dominant firms may mislead the researchers' judgments. The cross-sectional comparison of industrial clusters in different regions could be difficult. This paper demonstrates that industrial clusters can be recognized using principle component analysis (PCA). Based upon a correlation matrix, which is derived from the input-output table and measures backward and forward industrial linkages, this paper applies PCA to identifing industrial clusters. Several indices are further developed to evaluate PCA's performance in identifying regional industrial clusters. Correlation analysis is then applied to testing the spatial agglomeration of related industries. Taking Beijing as a case, this paper identifies 14 industrial clusters based on the 1997 inputing output table. Industrial clusters are formed around smelting and pressing in ferrous metals, chemicals, electronics, textiles and car production etc. Industries in those clusters are strongly tied with each other through sale-purchase linkages, and agglomerate in similar locales.