Use this panel to create clusters one by one until you are happy.
A. Specify parameters before searching for new clusters
Once the input data is processed and ready for clustering analysis, users need to specify a few parameters for clustering methods, so they can properly and quickly identify candidate clusters.
Number of seed: number of variables with the highest variance to be used as seeds of clusters; larger number will slow down the process; so number higher than 2,500 is not recommended.
Minimum size of clusters: the smaller number of variables in each cluster; so smaller clusters will not be reported.
Maximum number of clusters: maximum number of clusters to be reported; larger number will slow down the process; so number higher than 5 times of sample groups is not recommended.
Method: choose the clustering methods; methods currently available are hierarchical, k-means, and self-organizing map; more to come.
Evaluation: choose the methods to find the optimal number of clusters; methods currently available are silhouette, Dunn Index, and Davies-Bouldin's separation measure; more to come.
Remove orphan: orphans are variables don't have enough number (minimum cluster size - 1) of close neighbors; by default, they should be removed as they will contribute little to any cluster.
B. Select a candidate cluster
Clusters identified from the last step are summarized, so users can choose the one with the most interesting pattern to create a new cluster.
Summary statistics: the