Algorithm ranking based on results from both magnitude and shape cohorts.
<p>(A-B) and (C-D) show the average rank and Adjusted Rand Index (ARI), respectively, for all 30 algorithms across all cohorts. Average ranks were obtained by comparing all algorithm using Nemenyi tests in the R package <i>mlr3benchmark</i>, and lower average ranks indicate better...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | , , , , |
| Published: |
2024
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | <p>(A-B) and (C-D) show the average rank and Adjusted Rand Index (ARI), respectively, for all 30 algorithms across all cohorts. Average ranks were obtained by comparing all algorithm using Nemenyi tests in the R package <i>mlr3benchmark</i>, and lower average ranks indicate better performance. ARI scale ranges between -1 to 1, and values closer to zero represent classification on par with random assignment. (A) and (B) shows average ranks. Algorithms with similar accuracies are connected by black bars (dashed & solid) in (A). These metrics are further subset by missingness in (B), (C) and (D). (C) and (D) show ARI distributions for simulated datasets with no missingness and missingness <u>></u> 10%.</p> |
|---|