soiltechproject / fsm-docs

https://www.soiltechproject.org/
3 stars 0 forks source link

[Translation] k-means stability #11

Open KipCrossing opened 3 years ago

KipCrossing commented 3 years ago

Problem: The shape of the zones can change

I noticed that sometimes the the zones change with each run (same settings). eg: Run 1:

Screenshot from 2020-10-31 01-54-40

Run 2:

Screenshot from 2020-10-31 01-56-10

Note: the colour is of no concequence. look at the shape.

Solution: tl;dr - more itterations

The above example uses 10 iterations. By using 20 we get:

Screenshot from 2020-10-31 01-58-15

vs

Screenshot from 2020-10-31 01-59-01

Which is basically the same.

Question: will the size of the dataset change the stability?