Using data mining technique to explore anthropometric sizing stem data towards the development of sizing system

Abstract

An anthropometric survey of 629 girls aged between 7 and 12 years old were conducted covering major ethnic groups namely Malays. Chinese and Indians from schools in rural and urban districts of Selangor state in Malaysia. 33 different body dimensions were taken from each subject following the IS08559-1998 standard for body measurement. Firstly, the whole data was analysed using descriptive analysis of average, mean and standard deviation. The data was then further explored using the factor analysis method. Principal component analysis technique (PCA) was done to reduce the variables to similar factor components. Two components that have Eigen value more than 1 were selected. As a result, two important components were determined which is PC1 named as girth dimensions and PC2 is length dimensions. Next, two key variables were used to segment the children into 4 distinct clusters using K-means cluster method. Four body types were obtained from the segmentation known as stout, petite, slim and big groups. These cluster groups were validated using decision tree. These segmented groups will then be converted into size tables.