@randombishop
What I did is simply:
- Project my dataset of about 60 features per fid into 2 dimensions. (PCA1 and PCA2)
- Cluster the fids into 10 clusters using kmeans.
- Show the fids in 2D.
It doesn't say much so far, except there's a clear distinction between a big continent of users and a little island.
Maybe it will become more interesting when I start generating labels for the clusters and making the map interactive so you can colour the users by features. Like in this example, by "spamminess"
But I don't know yet, i am open to suggestions on where to take this approach...
0 reply
0 recast
0 reaction