Merge branch 'dev' of https://github.com/Quantmetry/qolmat into dev_test_rpca

Author: Gsaes

environment.ci.yml

@@ -12,4 +12,5 @@ dependencies:
           - numpydoc==1.5.0
           - pytest==7.2.0
           - pytest-cov==4.0.0
+          - tensorflow
           - -e .

examples/benchmark.md

@@ -329,7 +329,7 @@ dict_imputers["MLP"] = imputer_mlp = imputers_keras.ImputerRegressorKeras(estima
 
 We can re-run the imputation model benchmark as before.
 
-```python tags=[] jupyter={"outputs_hidden": true}
+```python tags=[]
 generator_holes = missing_patterns.EmpiricalHoleGenerator(n_splits=2, subset = cols_to_impute, ratio_masked=ratio_masked)
 
 comparison = comparator.Comparator(
@@ -350,7 +350,7 @@ plt.ylabel("mae")
 plt.show()
 ```
 
-```python jupyter={"outputs_hidden": true}
+```python
 df_plot = df_data
 dfs_imputed = {name: imp.fit_transform(df_plot) for name, imp in dict_imputers.items()}
 station = df_plot.index.get_level_values("station")[0]

qolmat/utils/data.py

@@ -189,7 +189,7 @@ def add_station_features(df: pd.DataFrame):
     """
     stations = df.index.get_level_values("station")
     for station in stations.unique():
-        df[f"station={station}"] = 1 * (stations == station)
+        df[f"station={station}"] = (stations == station).astype(float)
     return df
 
 
@@ -208,15 +208,9 @@ def add_datetime_features(df: pd.DataFrame):
         dataframe with missing values
     """
 
-    time = np.concatenate(
-        [
-            np.cos(2 * np.pi * np.arange(60, 366) / 365),
-            np.cos(2 * np.pi * np.arange(1, 366) / 365),
-            np.cos(2 * np.pi * np.arange(1, 366) / 365),
-            np.cos(2 * np.pi * np.arange(1, 367) / 366),
-            np.cos(2 * np.pi * np.arange(1, 60) / 365),
-        ]
+    time = df.index.get_level.values("datetime")
+    days_in_year = time.dt.year.apply(
+        lambda x: 366 if ((x % 4 == 0) and (x % 100 != 0)) or (x % 400 == 0) else 365
     )
-    for i_station, (station, dfs) in enumerate(df.groupby("station")):
-        df.loc[station, "Time"] = time
+    df["time_cos"] = np.cos(2 * np.pi * time.dt.dayofyear / days_in_year)
     return df