RapidMiner

RapidMiner

Decide for number of clusters

Contributor II

Decide for number of clusters

Hello,

I have extracted SSE, Avg. Sil width and Davies&Bouldin for 20 iterations on my data. The result table is below. I'm confused regarding choosing best k, since each method supports different k. Any idea/hint what should I do more?


                 



















kSSEASWDB
2 0.37492 0.89744 0.87625
3 0.19361 0.85002 1.37602
4 0.13999 0.76005 1.48418
5 0.07712 0.77031 1.60937
6 0.0579 0.76518 1.68158
7 0.03498 0.48512 1.84476
8 0.02939 0.57871 1.67001
9 0.01929 0.60316 1.62707
10 0.017 0.60372 1.61378
11 0.0175 0.60847 1.94819
12 0.01321 0.62244 1.66382
13 0.01197 0.63113 1.81006
14 0.01056 0.63378 1.80031
15 0.00845 0.64645 1.89968
16 0.00928 0.65362 1.69712
17 0.00618 0.65596 1.65225
18 0.00581 0.65596 1.59732
19 0.00605 0.65103 1.71599
20 0.00433 0.66539 1.68256


Kind regards,