Abstract: Based on upper garment′s detail sizes of lady′s suits,the samples are divided into four kinds: loose type,semi-loose type,semi-fit type and fit type by cluster analysis.Then characteristic indexes of upper garment structure of lady′s suit are extracted by using the methods of principal component analysis and correlation analysis.By adopting stepwise selection regression analysis,multivariate linear regression models are established.The expert knowledge of garment outline types,pattern characteristic indexes and their mathematic models of upper garment structure are obtained.The results provide theoretical basis for rational choice and determination of specifications in women′s garment structure design and lay a foundation for building auto-generation system of pattern size of the upper garment.