Optimal 'k'

What approach would you take to determine the optimal number of clusters (k) in k-means clustering?

Answer Panel