Graph-based data clustering via multiscale community detection

Liu, Zijing; Barahona, Mauricio

doi:10.1007/s41109-019-0248-7

Applied Network Science

Table 2 Comparison of graph constructions in terms of clustering performance (NMI and ARI) on eleven UCI datasets

From: Graph-based data clustering via multiscale community detection

Dataset	PMST	ε-ball	RMST k=1			kNN			CkNN δ=1
			γ=0.5	γ=0.25	γ=0.125	k=3	k=7	k=12	k=3	k=7	k=12
Normalised Mutual Information (NMI)
Iris	0.7764	0.8057	0.7106	0.7764	0.7400	0.7627	0.8057	0.8057	0.8226	0.7980	0.7777
Glass	0.3886	0.3862	0.3656	0.3646	0.3953	0.3863	0.3927	0.3626	0.3517	0.3941	0.4170
Wine	0.8080	0.7972	0.7389	0.8400	0.8633	0.7955	0.8336	0.8215	0.7528	0.8347	0.8113
WBDC	0.6042	0.5819	0.5839	0.6042	0.6182	0.6188	0.5944	0.6121	0.5822	0.7231	0.6113
Control chart	0.8272	0.8404	0.7602	0.7133	0.8130	0.8266	0.8520	0.852	0.8078	0.8883	0.8531
Parkinson	0.2220	0.2065	0.2012	0.2143	0.2811	0.3175	0.2815	0.2176	0.2113	0.2973	0.2423
Vertebral	0.5043	0.6273	0.5432	0.5323	0.5999	0.6060	0.5928	0.5424	0.6450	0.6018	0.6116
Breast tissue	0.5722	0.5616	0.5298	0.5820	0.5461	0.5447	0.5525	0.5243	0.5253	0.5648	0.5601
Seeds	0.7293	0.6943	0.7361	0.7570	0.6810	0.7515	0.7458	0.7318	0.6384	0.7381	0.7245
Image Seg.	0.6948	0.6605	0.6762	0.7072	0.7488	0.6154	0.6465	0.6667	0.6012	0.6347	0.6541
Yeast	0.2881	0.3051	0.2952	0.2626	0.2563	0.2764	0.2997	0.3080	0.2473	0.2959	0.3072
Average	0.5835	0.5879	0.5583	0.5776	0.5945	0.5910	0.5997	0.5859	0.5623	0.6155	0.5973
Adjusted Rand Index (ARI)
Iris	0.7420	0.7592	0.6603	0.7420	0.6957	0.7191	0.7592	0.7592	0.8184	0.7455	0.7445
Glass	0.2323	0.2099	0.1983	0.2029	0.2258	0.2278	0.2231	0.2266	0.2134	0.2398	0.2496
Wine	0.8350	0.8072	0.7375	0.8712	0.8823	0.8025	0.8498	0.8349	0.7414	0.8471	0.8360
WBDC	0.7193	0.7114	0.7014	0.7193	0.7369	0.7368	0.7010	0.7310	0.6697	0.8244	0.7200
Control chart	0.6929	0.7364	0.5694	0.5371	0.6991	0.6748	0.6824	0.7071	0.6902	0.8280	0.7227
Parkinson	0.2267	0.2205	0.2038	0.1540	0.2556	0.2176	0.2001	0.2045	0.1165	0.2667	0.2101
Vertebral	0.5257	0.6445	0.5702	0.5411	0.6015	0.5982	0.5802	0.5330	0.6441	0.6113	0.6302
Breast tissue	0.4689	0.4494	0.4100	0.4017	0.3494	0.4012	0.4272	0.4078	0.3631	0.3764	0.4471
Seeds	0.7353	0.7497	0.7458	0.7589	0.6687	0.7876	0.7889	0.7761	0.6402	0.7764	0.7655
Image Seg.	0.6060	0.5144	0.6030	0.6193	0.5942	0.3800	0.4669	0.5471	0.3791	0.4522	0.5121
Yeast	0.1908	0.2368	0.1827	0.1772	0.1649	0.1755	0.2230	0.2531	0.1797	0.1942	0.2294
Average	0.5432	0.5490	0.5075	0.5204	0.5340	0.5201	0.5365	0.5437	0.4960	0.5602	0.5516

A high NMI (and ARI) indicate that the best partition found by Markov Stability is similar to the ground truth, i.e., better clustering. The best performance for each dataset is marked in bold

Back to article page