Benchmarks

For benchmark our project we used the dataset provided by Marek Gagolewski. Specificallyh we are using wut dataset. Below you can see the table with our metrics of rand index and accuracy of our algorithms: kmeans and kmeans++ which was benchmarked on wut dataset. Also, we are providing visualization in order to give insights what data we used.

Comparison table

Test Case	Kmeans Rand Index	Kmeans Accuracy	Kmeans++ Rand Index	Kmeans++ Accuracy	Bkmeans Rand Index	Bkmeans Accuracy
circles	1.00	1.00	1.00	1.00	1.0	1.0
cross	0.5905	0.5967	0.5905	0.5967	0.7362	0.7167
graph	0.8835	0.2373	0.8852	0.6133	0.8687	0.3147
isolation	0.5564	0.3474	0.5557	0.3444	0.5543	0.3489
labirynth	0.7526	0.4511	0.7526	0.4511	0.7839	0.2350
mk1	1.00	1.00	1.00	1.00	1.0	1.0
mk2	0.5010	0.5467	0.5010	0.5467	0.5010	0.5467
mk3	0.9740	0.9778	0.9740	0.9778	0.9740	0.9778
mk4	0.5682	0.2311	0.5766	0.2400	0.6224	0.5111
olympic	0.7122	0.2907	0.7306	0.2707	0.7110	0.2840
smile	0.8263	0.4267	0.8379	0.5267	0.8335	0.5733
stripes	0.4998	0.5160	0.4998	0.5160	0.4998	0.5160
trajectories	1.00	1.00	1.00	1.00	1.0	1.0
trapped_lovers	0.5973	0.5840	0.5955	0.5813	0.5833	0.4533
twosplashes	0.6186	0.7500	0.6186	0.7500	0.6186	0.7500
windows	0.5745	0.3333	0.5677	0.3535	0.5820	0.4855
x1	1.00	1.00	1.00	1.00	1.0	1.0
x2	0.8301	0.8889	0.6078	0.6667	0.6078	0.6667
x3	0.9259	0.9286	0.9339	0.6429	0.9339	0.6429
z1	0.6256	0.4828	0.6256	0.4828	0.6650	0.5862
z2	0.7803	0.3333	0.7803	0.3333	0.7772	0.2963
z3	1.00	1.00	1.00	1.00	1.0	1.0

Visualization K-means

Figure 1: Clustering Circles with K-means

Figure 2: Mk1 with K-means

Figure 3: Mk3 with K-means

Figure 4: Z3 with K-means

Figure 5: Trajectories with K-means