Table 4.
Clustering performance obtained by the single controlled vocabulary and the multi-view approach
| Single CV | RI | NMI | Integration (9 CVs) | RI | NMI |
|---|---|---|---|---|---|
| LDDB | 0.7586 ± 0.0032 | 0.5290 ± 0.0032 | Ward linkage | 0.8236 ± 0 | 0.6015 ± 0 |
| OMIM | 0.7216 ± 0.0009 | 0.4606 ± 0.0028 | EACAL | 0.7741 ± 0.0041 | 0.5542 ± 0.0068 |
| Uniprot | 0.7130 ± 0.0013 | 0.4333 ± 0.0091 | OKKC(μmin = 0) | 0.7641 ± 0.0078 | 0.5395 ± 0.0147 |
| eVOC | 0.7015 ± 0.0043 | 0.4280 ± 0.0079 | MCLA | 0.7596 ± 0.0021 | 0.5268 ± 0.0087 |
| MPO | 0.7064 ± 0.0016 | 0.4301 ± 0.0049 | QMI | 0.7458 ± 0.0039 | 0.5084 ± 0.0063 |
| MeSH | 0.6673 ± 0.0055 | 0.3547 ± 0.0097 | OKKC(μmin = 1/N) | 0.7314 ± 0.0054 | 0.4723 ± 0.0097 |
| SNOMED | 0.6539 ± 0.0063 | 0.3259 ± 0.0096 | AdacVote | 0.7300 ± 0.0045 | 0.4093 ± 0.0100 |
| GO | 0.6525 ± 0.0063 | 0.3254 ± 0.0092 | CSPA | 0.7011 ± 0.0065 | 0.4479 ± 0.0097 |
| KO | 0.5900 ± 0.0014 | 0.1928 ± 0.0042 | Complete linkage | 0.6874 ± 0 | 0.3028 ± 0 |
| Average linkage | 0.6722 ± 0 | 0.2590 ± 0 | |||
| HGPA | 0.6245 ± 0.0035 | 0.3015 ± 0.0071 | |||
| Single linkage | 0.5960 ± 0 | 0.1078 ± 0 | |||
| Integration (9 LSI) | RI | NMI | Integration (35 subset CVs) | RI | NMI |
| Ward linkage | 0.7991 ± 0 | 0.5997 ± 0 | Ward linkage | 0.8172 ± 0 | 0.5890 ± 0 |
| OKKC(μmin = 0) | 0.7501 ± 0.0071 | 0.5220 ± 0.0104 | OKKC(μmin = 0) | 0.7947 ± 0.0052 | 0.5732 ± 0.0096 |
| EACAL | 0.7511 ± 0.0037 | 0.5232 ± 0.0075 | EACAL | 0.7815 ± 0.0064 | 0.5701 ± 0.0082 |
For integration of 9 LSI and 35 subset CVs, only the best three results are shown.