# Results
All additionl data can be found [here](https://sid.erda.dk/sharelink/EzFLHhjSI3)

## lsfml_benchmark
|  | Contains .log files for the crossvalidation
| `final_stats.job` | Contains the dictionary with all statistics for each run
| `stats_*.json.gz` | contains a dataframe of statistics for a single run. |

## no_duplicate_substrates
### Crossvalidation
| `logs` | Contains the logs for the crossvalidation |
| `metrics` | Contains combined metrics for the crossvalidation for all models |
| `per_model` | Contains model folders with metrics per scenario | 

## out_of_sample
| `literature_regio_yield_above_30` | Prediction results using LGBM (Opt) trained on the filtered dataset with top2 reagents and yield above 30 % |
| `no_duplicate_substrate` | Prediction results using LGBM (Opt) trained on the no_duplicate_substrate dataset |
| `top2_ir_precatalyst` | Prediction results using LGBM (Opt) trained on the top2_ir_precatalyst dataset |

## top2_ir_precatalyst
Contains subfolders for each feature set where:
| `iteration_*` | Folder containing results for each sub feature set, e.g. whether ligand encoding are present or not. |
| `results_*.json.gz` | Pandas dataframe with combined results for each sub feature set, e.g. whether ligand encoding are present or not. |
| `ablation_study_results.parquet` | parquet file with all results for each model and sub feature set, e.g. whether ligand encoding are present or not. |