Table 3. The clustering accuracy of different models on the correct number of topics.
The best results are highlighted in bold (The higher the better).
Datasets | Topic models | Metrics | ||
---|---|---|---|---|
NMI | ARI | ACC | ||
BBCsport | LDA | 0.709 | 0.662 | 0.851 |
NMF | 0.818 | 0.856 | 0.872 | |
K-Means | 0.894 | 0.896 | 0.963 | |
BBCNews | LDA | 0.727 | 0.701 | 0.862 |
NMF | 0.812 | 0.842 | 0.932 | |
K-Means | 0.751 | 0.726 | 0.868 | |
Reuters | LDA | 0.432 | 0.443 | 0.669 |
NMF | 0.550 | 0.597 | 0.776 | |
K-Means | 0.632 | 0.671 | 0.832 | |
AGNews | GSDMM | 0.585 | 0.622 | 0.833 |
SeaNMF | 0.563 | 0.600 | 0.822 | |
Snippets-1 | GSDMM | 0.565 | 0.592 | 0.830 |
SeaNMF | 0.580 | 0.634 | 0.850 | |
Snippets-2 | GSDMM | 0.764 | 0.823 | 0.919 |
SeaNMF | 0.787 | 0.850 | 0.939 |