add benchmark filtering for problem type

Author: paulbkoch

docs/benchmarks/ebm-benchmark.ipynb

@@ -912,7 +912,7 @@
     "print(f'Results (pre-filtered) count: {results_df.shape[0]}')\n",
     "\n",
     "# Optionally filter out results we want to replace\n",
-    "#results_df = results_df[~(results_df['method'] == 'ebm')]\n",
+    "#results_df = results_df[results_df['method'] != 'ebm']\n",
     "#results_df = results_df[~((results_df['method'] == 'ebm') & (results_df['meta'] == '{}'))]\n",
     "print(f'Results (post-filtered) count: {results_df.shape[0]}')"
    ]
@@ -949,18 +949,19 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Optionally filter out any incomplete datasets\n",
-    "#results_df = results_df[~(results_df['task'] == 'Devnagari-Script')]\n",
-    "print(f'Final count: {results_df.shape[0]}')\n",
-    "\n",
-    "\n",
     "types_df = results_df[results_df['name'].isin(['auc', 'ovo_auc', 'nrmse'])]\n",
     "task_to_type = types_df.groupby('task')['name'].first().map({'auc': 'binary', 'ovo_auc': 'multiclass', 'nrmse': 'regression'})\n",
     "results_df['type'] = results_df['task'].map(task_to_type).fillna('')\n",
     "\n",
     "flip = ['r2', 'auc', 'precision', 'recall', 'accuracy', 'bal_acc', 'ovo_auc', 'ovr_auc', 'mprecision', 'mrecall', 'maccuracy', 'mbal_acc']\n",
     "condition = results_df['name'].isin(flip)\n",
-    "results_df.loc[condition, 'num_val'] = -results_df.loc[condition, 'num_val']"
+    "results_df.loc[condition, 'num_val'] = -results_df.loc[condition, 'num_val']\n",
+    "\n",
+    "\n",
+    "# Optionally filter out any incomplete datasets\n",
+    "#results_df = results_df[results_df['task'] != 'Devnagari-Script']\n",
+    "#results_df = results_df[results_df['type'] == 'regression']\n",
+    "print(f'Final count: {results_df.shape[0]}')"
    ]
   },
   {