changed metric computation to suppress future warnings, removed verbo…

…se copula fitting message

pyforecaster/copula.py

@@ -11,7 +11,7 @@ class GaussianCopula:
     def __init__(self, cov_est_method='vanilla', logger=None):
         self.pars = None
         self.cov_est_method = cov_est_method
-        self.logger = logger if logger is not None else get_logger(name='Copula')
+        self.logger = logger if logger is not None else get_logger(name='Copula', level=logging.WARNING)
         self.dim = None
 
     def fit(self, y, x=None, do_plot=True):

pyforecaster/metrics.py

@@ -17,18 +17,36 @@ def squerr(x, t):
 
 def rmse(x, t, agg_index=None, **kwargs):
     agg_index = np.ones_like(x.index) if agg_index is None else agg_index
-    return squerr(x, t).groupby(agg_index, axis=chose_axis(x, agg_index)).mean() ** 0.5
+    res = squerr(x, t)
+    axis = chose_axis(x, agg_index)
+    if axis == 0:
+        res = res.groupby(agg_index).mean() ** 0.5
+    else:
+        res = res.T.groupby(agg_index).mean() ** 0.5
+    return res
 
 
 def mape(x, t, agg_index=None, **kwargs):
     agg_index = np.ones_like(x.index) if agg_index is None else agg_index
-    return (err(x, t)/(t + 1e-5)).abs().groupby(agg_index, axis=chose_axis(x, agg_index)).mean()
+    res = (err(x, t)/(np.abs(t) + 1e-5)).abs()
+    axis = chose_axis(x, agg_index)
+    if axis == 0:
+        res = res.groupby(agg_index).mean()
+    else:
+        res = res.T.groupby(agg_index).mean()
+    return res
 
 
 def nmae(x, t, agg_index=None, inter_normalization=True, **kwargs):
     agg_index = np.ones_like(x.index) if agg_index is None else agg_index
     offset = t.abs().mean(axis=1).quantile(0.5) * 0.01 + 1e-12 if inter_normalization else 1e-12
-    return (err(x, t) / (t.abs().mean(axis=1).values.reshape(-1,1) + offset)).abs().groupby(agg_index, axis=chose_axis(x, agg_index)).mean()
+    axis = chose_axis(x, agg_index)
+    res = (err(x, t) / (t.abs().mean(axis=1).values.reshape(-1,1) + offset)).abs()
+    if axis == 0:
+        res = res.groupby(agg_index).mean()
+    else:
+        res = res.T.groupby(agg_index).mean()
+    return res
 
 
 def quantile_scores(q_hat, t, alphas=None, agg_index=None, **kwargs):

pyforecaster/scenarios_generator.py

@@ -2,6 +2,7 @@
 from scipy.interpolate import interp1d
 from .scenred import superimpose_signal_to_tree
 from pyforecaster.utilities import get_logger
+from logging import WARNING
 import pandas as pd
 import pyforecaster.dictionaries
 import numpy as np
@@ -17,7 +18,7 @@ class ScenGen:
     def __init__(self, copula_type: str = 'HourlyGaussianCopula', tree_type:str = 'ScenredTree',
                  online_tree_reduction=True, q_vect=None, nodes_at_step=None, max_iterations=100,
                  parallel_preds=False, additional_node=False, prefit_trees=False, **kwargs):
-        self.logger = get_logger()
+        self.logger = get_logger(level=WARNING)
         self.copula_type = copula_type
         self.tree_type = tree_type
         self.copula_class = pyforecaster.dictionaries.COPULA_MAP[copula_type]