Figure 7
From: Finding reproducible cluster partitions for the k-means algorithm
Performance chart for 10,000 dataset. Ordered Cramérs' V for 100 runs of the SeCo framework with selection of the best solution being (a) SSQ only and (b) Median Internal Cramérs' V for the top 10% of solutions. Results for the dataset with 10,000 data points. Red lines highlights k = 10, blue lines k = 9 and the black lines k = 8. Confidence intervals for k = 8 (Upper: +0.007, Lower: -0.008) k = 9 (Upper: +0.006, Lower: -0.007) k = 10 (Upper: +0.006, Lower: = -0.007).