Table 5. The impact of low-frequency words on the performance of AICDR.
The best results are highlighted in bold.
| Datasets | Topic models | Methods | ||
|---|---|---|---|---|
| One-Hot | BOW | TF-IDF | ||
| BBCsport | LDA | 5 | 5 | 5 |
| NMF | 4 | 4 | 4 | |
| K-Means | 5 | 2 | 5 | |
| BBCNews | LDA | 6 | 6 | 6 |
| NMF | 5 | 5 | 5 | |
| K-Means | 3 | 3 | 6 | |
| Reuters | LDA | 3 | 3 | 3 |
| NMF | 3 | 3 | 3 | |
| K-Means | 3 | 2 | 3 | |
| AGNews | GSDMM | 4 | 4 | 4 |
| SeaNMF | 5 | 5 | 5 | |
| Snippets-1 | GSDMM | 4 | 4 | 4 |
| SeaNMF | 4 | 4 | 4 | |
| Snippets-2 | GSDMM | 3 | 3 | 3 |
| SeaNMF | 3 | 3 | 3 | |