With the improvement of material living standards, consumers have a higher demand for fitting clothing. Due to the characteristics of low cost, high quality, and fast speed, mass customization of clothing has become an important way to meet the needs of consumers for high fitness and good production efficiency of enterprises. As the basis of mass customization of clothing, clothing size is prone to mismatches between size and individual body shape, making it difficult for clothing to meet consumers’ higher fitness needs. Therefore, the technical methods for formulating clothing sizes should be innovated to further meet the needs of mass customization of clothing. To meet the higher demand of consumers for clothing fit and the need of enterprises to improve the production efficiency of clothing customization, a method for formulating customized sizes for young women’s clothing considering the loss coefficient of fit was proposed. Eight human body data were collected from 100 young women aged between 18 and 25 by using [TC]~2 three-dimensional human body scanning. The measurement sites included height, chest circumference, waist circumference, hip circumference, back length, shoulder width, leg length, and thigh circumference. After performing data preprocessing on the collected initial human body data, we deleted the missing and abnormal values, and then performed a normal distribution test. It is verified that the data conform to normal distribution. We also performed principal component analysis on the pre-processed data to determine the number of components, and extracted two components with initial feature values greater than 1. After determining the basis variables for clustering through correlation analysis, the optimal number of clusters was determined by using the loss coefficient of fit method. Then, the sample data were classified through K-means clustering analysis, and the linear regression analysis was used to set the grade to obtain a customized size table. To prove the superiority of the loss coefficient of the fit method, two other methods including the mixed F statistics and the sum of final cluster center distance for determining the number of clusters were selected, and a comparative experiment was conducted. According to this research idea, we established a loss coefficient of fit model and selected the optimal cluster number from the perspective of clothing and individual fitting to determine the optimal size table for clothing mass customization. The research results show that the coverage rate of the clothing size table based on the loss coefficient of the fit method is as high as 95. 40%, and the fitting loss is only 0. 039 9. It is superior to the mixed F statistics and the sum of the final clustering center distance in terms of size coverage, clothing fitness, and production efficiency. The method of the loss coefficient of fit provides a new way of thinking for the formulation of clothing size, and improves the size coverage rate based on giving priority to meeting the fitness needs of consumers. The research results can provide customized size development methods for single orders of niche brands and small and medium-sized enterprises. It is unnecessary to build a three-dimensional human body database for consumers, but necessary to quickly develop a clothing mass customization size table by collecting data on the eight important measurement positions of consumers within the order. [ABSTRACT FROM AUTHOR]