Non-hierarchical cluster analysis is the next step to a hierarchical cluster model. It allows the partitioning of the similar matrices into equal numbers of clusters. It also creates a list of the partitions from the similar matrix generated in the hierarchical cluster. Each partition comprises of at least one item. The previous article explained the steps to hierarchical clustering and its interpretations were presented. Furthermore, this article will present an explanation on how to perform non-hierarchical clustering in Hamlet II.
Non-hierarchical cluster analysis
Creation of joint frequencies helps in the formation of both hierarchical and non-hierarchical clusters. After saving the matrix of similarities from the hierarchical clusters separately, the next step is to conduct a non-hierarchical cluster analysis.
Follow these steps in order to conduct non-hierarchical clustering in Hamlet II as shown in the figure below:
- Click on “Cluster Analysis”.
- Select the option “non-hierarchical” from the drop-down menu.
- Select the previously saved matrix (*.mat) file from the destination folder.
- Click on ‘Open’.
The results for the test will appear.
Interpreting the result
This process established an upper bound with the help of biased-sampling link cluster analysis and Branch-and-Bound clustering method. Thus, it minimized the partition diameter in order to separate the matrix containing similarities into clusters. As the above figure shows, there were 2 – 9 partitions in 8 iterations until the minimum diameter emerged. Furthermore, a solution set with the least minimum diameter (closest to 1) is ideal for best results.
Non-hierarchical cluster analysis simply creates a list containing all the partitions generated acquired from the mentioned finding in hierarchical clusters. If there are less than or equal to 15 entries then Hamlet II automatically lists all the cases of segments from 2 to the total number of elements minus 1.
This article focuses on the text-based non-hierarchical cluster analysis using Hamlet II with an example. In the next article, correspondence analysis is presented for multiple file comparison and to represent row and column variables in the same space using singular-value decomposition.