How can I identity clusters on my scatterplot?

So... I plotted the following scatterplot. There is clearly a correlation between the 3 top vehicle categories (B-SPORT, D-SDN and D-SUV). I need to make a cluster that represents this correlation and the correlation between the other 7 low categories. Is there a way to do this?

Answer 1

[Image Analyst]

It looks like you don't have clusters. And I doubt you used scatter() to plot those. It looks like you used plot() to plot 10 separate sets of (x,y) data. Hence each set is already it's own "cluster", if you want to call it that, and shows up in its unique color. So I don't think you need to find clusters in the points like you said in your subject line. But you may want to find correlations between each data set and all the other 9 data sets, like you said in the body of your post. To do that you can use corrcoef.

Answer 2

[Sulaymon Eshkabilov]

There are a few options in matlab for data clustering.

(1) using clusterdata() function: SEE

(2) using findcluster tool: SEE

(3) using fuzzy c-means clustering: SEE

[Brenno Selli]

I tried them all, the second one seems to work best but it's not what I'm looking for. In all clusters the scatter does not show the data so that I can see which category has similarity with another. I need to know this specifically, so then I plot separately showing it better.

[Sulaymon Eshkabilov]

An alternative data clustering tool or fcn of Statistics and ML toolbox is using spectralcluster()