Abstract: In order to obtain soybean varieties than can be specially used for soybean milk production, principal component analysis (PCA) and cluster analysis (CA) were used to screen 29 soybean varieties in Heilongjiang. Firstly, the physicochemical properties (contents of water, fat, soluble sugar, protein, calcium and protein subunit 7S and 11S, and 11S/7S ratio) of soybean and the quality indexes of soybean milk (protein content, stability, yield, particle size, protein transfer rate, and sensory score) were analyzed. Then, by analyzing the obtained data using PCA, the first six principal components (comprehensive evaluation indicators) were selected, explaining 84.97% of the total variance. The top three varieties with the highest comprehensive scores were ‘Heinong 85’, ‘Heinong 66’ and ‘Heinong 71’, whose comprehensive scores were significantly higher than those of other varieties (P < 0.05). Finally, the results of cluster analysis showed that the 29 varieties of soybean milk were divided into categories I, II, III and IV, consisting of 14, 11, 3 and 1 varieties, respectively. The three varieties selected by PCA were all in category II, which had the characteristics of high protein content, protein transfer rate, and sensory score as well as medium soymilk yield, meeting the technical requirements of soymilk products. In conclusion, ‘Heinong 85’, ‘Heinong 66’ and ‘Heinong 71’ can be specially used for soybean milk production, and PCA and CA are suitable for screening and comprehensive evaluation of multiple samples and multiple indicators.

Key words: soybean milk, principal component analysis, cluster analysis, quality evaluation