| Metric | Train | Validation | Test |
|---|---|---|---|
| accuracy | 0.9234 | 0.8912 | 0.8856 |
| f1 | 0.9201 | 0.8876 | 0.8823 |
| precision | 0.9156 | 0.8845 | 0.8798 |
| recall | 0.9245 | 0.8907 | 0.8849 |
| roc_auc | 0.9789 | 0.9543 | 0.9512 |
| log_loss | 0.2134 | 0.2876 | 0.3012 |
| Key | Value |
|---|---|
| Predictor type | MultiModalPredictor |
| Framework | AutoGluon Multimodal |
| Model architecture | timm_image=resnet50, hf_text=bert-base-uncased |
| Modalities & Inputs | Images + Tabular |
| Label column | target |
| Image columns | image_path |
| Tabular columns | 15 |
| Presets | medium_quality |
| Eval metric | accuracy |
| Decision threshold calibration | enabled |
| Decision threshold (Test only) | 0.500 |
| Seed | 42 |
| time limit(s) | 3600 |
| Class | Count | Percent |
|---|---|---|
| 0 | 1245 | 45.23% |
| 1 | 1508 | 54.77% |
| Metric | Train | Validation |
|---|---|---|
| accuracy | 0.9234 | 0.8912 |
| f1 | 0.9201 | 0.8876 |
| precision | 0.9156 | 0.8845 |
| recall | 0.9245 | 0.8907 |
| roc_auc | 0.9789 | 0.9543 |
| log_loss | 0.2134 | 0.2876 |
| Metric | Test |
|---|---|
| accuracy | 0.8856 |
| f1 | 0.8823 |
| precision | 0.8798 |
| recall | 0.8849 |
| roc_auc | 0.9512 |
| log_loss | 0.3012 |
| specificity (TNR) | 0.8765 |
| sensitivity (Sensitivity/TPR) | 0.8923 |
Permutation importance is not supported for MultiModalPredictor in this tool. For tabular-only runs, this section shows permutation importance.
This run used MultiModalPredictor (images + tabular).
Label column: target
Image column: image_path