Abstract: 16 related indicators were investigated using 130 Gossypium barbadense hybrids as experimental materials, the relationship among main agronomic traits, yield and fiber quality was explored through correlation, principal component and regression analysis, and excellent hybrid materials were screened by systematic clustering. The results showed that lint yield and seed cotton yield were positively correlated with plant height, fruit branches, effective branches and bell number, and negatively correlated with the height of the initial section and the number of the first section, while there was a mutual constraint relationship between fiber quality and yield and fiber quality components. After dimensionality reduction of 16 indexes, five principal components were extracted, named fruit branch factor, yield factor, upper average length and uniformity determinant, fiber quality factor and growth factor, which explained 82.16% of the variation of the population. Seven optimal regression models were constructed with yield traits and quality traits as dependent variables and main agronomic traits as independent variables, which provided a basis for the rapid and accurate selection of group materials to meet different needs. Yield clustering and quality clustering of the test materials were carried out, and 6 high-yield and 10 high-quality hybrid combinations were screened respectively, and hybrid breeding provided the possibility for genetic information recombination, and the specific selection of parents was expected to achieve this breeding goal.