"how to programatically cluster a vector into X, Y coordinates?"

daviddonohue

Hello!  I would like to convert a table of data, with 5 or more attributes, into a set of X, Y coordinates, suitable for plotting in my own scatter plot.  I would think that I could represent each row as a vector of numeric and nominal values.  Next, I could pass the vectors into a RapidMiner clustering algorithm, and retrieve a single pair of X,Y coordinates for each row.  Can RapidMiner help me do this?
Many thanks!
David Donohue

land

Hi David,
RapidMiner can help here, but not with clustering. Clustering might provide some informations about centroids within groups, but the coordinates of the centroids are in the space of the data and hence have the same dimensionality. But you might use algorithms for reducing the dimensionality. I would suggest using the PCA, because you have to build the model only once and then might apply it to any new data, placing it in the same way. If you use SOMDimensionalityReduction, you will have to retrain the complete model everytime you have new data.

Greetings,
  Sebastian