Analysis of clustering by decision treeThere are two types of cluster analysis , hierarchical and non-hierarchical. In the hierarchical type, the results are displayed in a graph called a dendrogram, which can be used to proceed with the analysis.
However, there is no standard graph for non-hierarchical types. If it is 2D data, you can see how the grouping was done in the 2D scatter plot , but if it is 3D or more, you will not know.
This page is a countermeasure.
You can use visualization to compress high dimensions into two dimensions to convert multidimensional data into two-dimensional data. At this time, multidimensional features become visible in 2D.
Once it becomes two-dimensional, the problem at the beginning will be solved.
There is a method of compressing the high dimension to 2D and color-coding the result of the 2D map by cluster analysis on the page of visualization and self-organizing map .
In the above method, the data handled by the cluster analysis is converted into two dimensions, so the ability to handle data in three or more dimensions by the cluster analysis method is not used.
One way to use this ability is to make only the data for the graph two-dimensional.
The method on this page is more suitable when the sample size is small. Also, the method on this page doesn't tell you "Why did clustering happen?"
A way to compensate for these weaknesses is to analyze clustering with a decision tree .
NEXT Analysis of clustering by decision tree
Tweet