Page 72 - Anthropometry, Apparel Sizing and Design

Developing apparel sizing system using anthropometric data 101

4.2.6 Step 6: Principal component analysis (PCA)

The objective of using PCA is to reduce the number of variables and to cluster these
variables into a more parsimonious and manageable number of groups. Parsimonious
means to summarize most of the original information (variance) in a minimum num-
ber of components for prediction purposes (Pallant, 2001).

4.2.7 Step 7: Cluster analysis

Cluster analysis is an exploratory data analysis tool used to segment a population into
homogenous subgroups. This means that each person in a group shares similar phys-
ical traits with others in the group and that people in one group differ from those in
other groups.

4.2.8 Step 8: Classification analysis (decision tree)

Decision tree analysis is a data mining technique that is effective for classification
(Lin et al., 2008). The classification and regression tree (CRT) technique can be used
to verify and classify the sample population according to cluster groups; CRT is used
where the data are continuous. The profile of the tree is useful when interpretation of
the data set is required. By doing the classification analysis, important variables can be
obtained, and a simple profile can easily be extracted from the tree diagram (Viktor
et al., 2006).
The last stage described in Fig. 4.1 is Stage 3—the sizing system development.

4.2.9 Step 9: Size system development

The purpose of developing the sizing system is to create sizes for each cluster group
that are appropriate to the individual group’s range. Two important decisions must be
made. The first is to estimate the size roll, which will accommodate most of the target
population, and the second is to determine which samples go into the cluster groups
obtained from the cluster analysis technique (Bairi et al., 2017). The goal is to accom-
modate as many people from the target population as possible using one intersize
interval.
For the development of the sizing system, the following elements have to be cal-
culated: size range, size interval, size scale, and size roll.
After the selection of the interval range, the classification profile obtained from the
decision tree analysis is used as a guide to select samples matching the right body size
and shapes. Using this profile the samples are classified according to the body sizes
and shapes. The last step is to validate the efficiency and accuracy of the sizing system
thus developed.