[BUG] Added level domain protection for normality and permbu me…

…thods. (#166) * Added mypy/flake8 installation to CONTRIBUTING instructions * Added level (0,100] domain protection Co-authored-by: fede <[email protected]>
Nixtla · Jan 23, 2023 · 2eadc28 · 2eadc28
1 parent 7cce341
commit 2eadc28
Show file tree

Hide file tree

Showing 4 changed files with 47 additions and 10 deletions.
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -60,7 +60,7 @@ nbdev_export
 ```
 
 ### Check syntax with Linters
-This project uses a couple of linters to validate different aspects of the code. Before opening a PR, please make sure that it passes all the linting tasks by following the next steps.
+This project uses a couple of linters to validate different aspects of the code. Before opening a PR, please make sure that it passes all the linting tasks by following the next steps. After installing `pip install flake8` and `pip install mypy`.
 
 * `mypy hierarchicalforecast/`
 * `flake8 --select=F hierarchicalforecast/`

diff --git a/hierarchicalforecast/_modidx.py b/hierarchicalforecast/_modidx.py
@@ -180,4 +180,4 @@
                                             'hierarchicalforecast.utils.is_strictly_hierarchical': ( 'utils.html#is_strictly_hierarchical',
                                                                                                      'hierarchicalforecast/utils.py'),
                                             'hierarchicalforecast.utils.numpy_balance': ( 'utils.html#numpy_balance',
-                                                                                          'hierarchicalforecast/utils.py')}}}
+                                                                                          'hierarchicalforecast/utils.py')}}}
diff --git a/hierarchicalforecast/core.py b/hierarchicalforecast/core.py
@@ -93,8 +93,9 @@ def _prepare_fit(self,
                      S_df: pd.DataFrame,
                      Y_df: Optional[pd.DataFrame],
                      tags: Dict[str, np.ndarray],
-                     intervals_method: str, 
-                     sort_df: bool):
+                     level: Optional[List[int]] = None,
+                     intervals_method: str = 'normality',
+                     sort_df: bool = True):
         """
         Performs preliminary wrangling and protections
         """
@@ -122,7 +123,13 @@ def _prepare_fit(self,
 
         if self.insample or (intervals_method in ['bootstrap', 'permbu']):
             if Y_df is None:
-                raise Exception('you need to pass `Y_df`')            
+                raise Exception('you need to pass `Y_df`')
+
+        # Protect level list
+        if (level is not None):
+            level_outside_domain = np.any((np.array(level) <= 0)|(np.array(level) > 100 ))
+            if level_outside_domain and (intervals_method in ['normality', 'permbu']):
+                raise Exception('Level outside domain, send `level` list in (0,100]')
 
         # Declare output names
         drop_cols = ['ds', 'y'] if 'y' in Y_hat_df.columns else ['ds']
@@ -186,7 +193,7 @@ def reconcile(self,
         If a class of `self.reconciles` receives `y_hat_insample`, `Y_df` must include them as columns.<br>
         `S`: pd.DataFrame with summing matrix of size `(base, bottom)`, see [aggregate method](https://nixtla.github.io/hierarchicalforecast/utils.html#aggregate).<br>
         `tags`: Each key is a level and its value contains tags associated to that level.<br>
-        `level`: float list 0-100, confidence levels for prediction intervals.<br>
+        `level`: positive float list (0-100], confidence levels for prediction intervals.<br>
         `intervals_method`: str, method used to calculate prediction intervals, one of `normality`, `bootstrap`, `permbu`.<br>
         `num_samples`: int=-1, if positive return that many probabilistic coherent samples.
         `seed`: int=0, random seed for numpy generator's replicability.<br>
@@ -201,6 +208,7 @@ def reconcile(self,
                                       S_df=S,
                                       Y_df=Y_df,
                                       tags=tags,
+                                      level=level,
                                       intervals_method=intervals_method,
                                       sort_df=sort_df)
 

diff --git a/nbs/core.ipynb b/nbs/core.ipynb
@@ -193,8 +193,9 @@
     "                     S_df: pd.DataFrame,\n",
     "                     Y_df: Optional[pd.DataFrame],\n",
     "                     tags: Dict[str, np.ndarray],\n",
-    "                     intervals_method: str, \n",
-    "                     sort_df: bool):\n",
+    "                     level: Optional[List[int]] = None,\n",
+    "                     intervals_method: str = 'normality',\n",
+    "                     sort_df: bool = True):\n",
     "        \"\"\"\n",
     "        Performs preliminary wrangling and protections\n",
     "        \"\"\"\n",
@@ -222,7 +223,13 @@
     "\n",
     "        if self.insample or (intervals_method in ['bootstrap', 'permbu']):\n",
     "            if Y_df is None:\n",
-    "                raise Exception('you need to pass `Y_df`')            \n",
+    "                raise Exception('you need to pass `Y_df`')\n",
+    "        \n",
+    "        # Protect level list\n",
+    "        if (level is not None):\n",
+    "            level_outside_domain = np.any((np.array(level) <= 0)|(np.array(level) > 100 ))\n",
+    "            if level_outside_domain and (intervals_method in ['normality', 'permbu']):\n",
+    "                raise Exception('Level outside domain, send `level` list in (0,100]')\n",
     "\n",
     "        # Declare output names\n",
     "        drop_cols = ['ds', 'y'] if 'y' in Y_hat_df.columns else ['ds']\n",
@@ -286,7 +293,7 @@
     "        If a class of `self.reconciles` receives `y_hat_insample`, `Y_df` must include them as columns.<br>\n",
     "        `S`: pd.DataFrame with summing matrix of size `(base, bottom)`, see [aggregate method](https://nixtla.github.io/hierarchicalforecast/utils.html#aggregate).<br>\n",
     "        `tags`: Each key is a level and its value contains tags associated to that level.<br>\n",
-    "        `level`: float list 0-100, confidence levels for prediction intervals.<br>\n",
+    "        `level`: positive float list (0-100], confidence levels for prediction intervals.<br>\n",
     "        `intervals_method`: str, method used to calculate prediction intervals, one of `normality`, `bootstrap`, `permbu`.<br>\n",
     "        `num_samples`: int=-1, if positive return that many probabilistic coherent samples.\n",
     "        `seed`: int=0, random seed for numpy generator's replicability.<br>\n",
@@ -301,6 +308,7 @@
     "                                      S_df=S,\n",
     "                                      Y_df=Y_df,\n",
     "                                      tags=tags,\n",
+    "                                      level=level,\n",
     "                                      intervals_method=intervals_method,\n",
     "                                      sort_df=sort_df)\n",
     "\n",
@@ -910,6 +918,27 @@
     "bootstrap_df"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "#| hide\n",
+    "# test level protection for PERMBU and Normality probabilistic methods\n",
+    "hrec = HierarchicalReconciliation([BottomUp()])\n",
+    "test_fail(\n",
+    "    hrec.reconcile,\n",
+    "    contains='Level outside domain',\n",
+    "    args=(hier_grouped_hat_df, S_grouped_df, tags_grouped, hier_grouped_df, [0, 80, 90], 'permbu',)\n",
+    ")\n",
+    "test_fail(\n",
+    "    hrec.reconcile,\n",
+    "    contains='Level outside domain',\n",
+    "    args=(hier_grouped_hat_df, S_grouped_df, tags_grouped, hier_grouped_df, [80, 90, 101], 'normality',)\n",
+    ")"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},