With the improvement of social and economic level, people have higher and higher requirements for the comfort and fit of clothing, and the body type classification of the national size standard can no longer accurately summarize the characteristics of the human body nowadays, which leads to the poor fit of clothing and the human body. At the same time, the competition in the clothing industry is intensifying, and the development of personalized clothing customization requires a more detailed classification of local human body types. Buttock morphology is an important factor that affects whether the structural design of underwear fits and is comfortable, and a reasonable classification of buttock shape is very helpful for the structural design of underwear. In this project, young women are selected as the research object, to make up for the lack of age division of the national standard size, and the human body data of 125 young women in school (Wuhan Textile University) are measured by three-dimensional human body scanner, and hipline, hip depth, waist-hip angle, and hip angle are taken as the classification indexes on the basis of the principal component factor analysis method and correlation analysis. In order to improve the accuracy of the classification of body features, the optimal number of categories was determined by using the contour coefficient, and the young women's human buttock morphology was subdivided into six categories on the basis of the Self-Organizing Feature Mapping Network (SOM) clustering model. Finally, Matlab software was used to construct a discriminate model of young women's buttock morphology based on Long Short-Term Memory Network (LSTM), and the results showed that the classification accuracy of the model was as high as 100%, and compared with BP neural network, PNN neural network, and GRNN neural network, the operation status, fitting effect, and accuracy rate of the model were higher than those of BP neural network, PNN neural network,...