Optimal clustering methods
There are many ways to determine what is the best separation of the clusters. In this package the function OptimalGeneSets can help with determining what is the best separation.
There are 3 different statistics the function can plot that help with this; Gap, Elbow and Silhouette.
- Gap: Compares the total within intra-cluster variation for different values of k with their expected values under null reference distribution of the data
- Elbow: For each k, calculate the total within cluster sum of squares
- Silhouette: Determines how wel each object lies within its cluster. Higher the better.
The computational time for each method is different:
for an average of 10 iterations of example data the time it took R to calulate.
Mean times:
- Gap: 23.35 ± 2.88
- Elbow: 0.76 ± 0.174
- Silhouette: 0.75 ± 0.158