3.2. Cluster analysis
Using the hierarchical clustering method, with the clustering step being set up at one level, 12 clusters (named Cluster 1
to Cluster 12) were obtained. Each of these clusters could be regarded as a research theme. The 150 keywords from the normalization
process were divided into 12 theme-clusters. If keywords are grouped into a cluster, they are more likely to have
an identical research theme. In this study, five to ten keywords with top frequencies and co-word data (co-word frequency
and correlation coefficient) were selected to represent each theme-cluster, because these keywords were most likely to be
chosen and used by researchers in each cluster. Table 3 shows the representative keywords within each theme-cluster.