【objective】 The main purpose of the present study was to accurately evaluate soybean germplasm resources, build the comprehensive evaluation system, and satisfy the needs of fast and direct selection of breeding parent for breeder. 【method】 11 important agronomic traits in 451 soybean germplasm resources of Yunnan province, were studied by path analysis, principal component analysis, membership function method, clustering analysis, and stepwise regression analysis. 【result】 The results showed that there were plentiful phenotypic variations and the complex correlation among the 11 important agronomic traits. And the mutual influence between the 11 traits was clearly divided into two paths, one final impact factor was the quality of soybean, another final impact factor was seed weight per plant. The 11 single indexes were transformed into 5 independent comprehensive components through principal component analysis, which represent 87.685% variation of the raw data. The membership function method was employed to calculate comprehensive value (D) and carried on cluster analysis, and 451 soybean germplasm resources were divided into three types. 97 soybean germplasm resources of the best comprehensive performance wereSfiltered. At last, a mathematical evaluation model for soybean germplasm resources was established by stepwise regression analysis, _。and its accuracy was higher than 95.25%. 【conclusion】 97 soybean germplasm resources of the best comprehensive performance wereSfiltered to form 451 soybean germplasm resources, and we construct a comprehensive evaluation system by using 11 traits.