Forest CoverType: SVM vs Random Forest
Random Forest and SVM compared on a multiclass problem, with cross-validation and per-class metrics.
What was built
Exploratory analysis first, then Random Forest with different tree counts and SVMs with C from 0.1 to 1000. SVM accuracy rises from 0.721 to 0.833 over that range. The 5-fold cross-validated weighted F1 is 0.776.
Facts
| Item | Value |
|---|---|
| SVM accuracy, C 0.1 to 1000 | 0.721 to 0.833 |
| SVM 5-fold weighted F1 | 0.776 |
Links
No public repository for this project.