1  import pandas as pd
 2  import sklearn.metrics
 3  
 4  use_new_root_mean_squared_error = hasattr(sklearn.metrics, "root_mean_squared_error")
 5  
 6  
 7  def root_mean_squared_error_compat(y_true, y_pred):
 8      """
 9      Compute the Root Mean Squared Error (RMSE) in a way that is compatible
10      with both old and new versions of scikit-learn.
11  
12      In scikit-learn >= 1.6.0, uses sklearn.metrics.root_mean_squared_error.
13      In earlier versions, uses mean_squared_error with squared=False.
14      """
15      if use_new_root_mean_squared_error:
16          from sklearn.metrics import root_mean_squared_error
17  
18          return root_mean_squared_error(y_true, y_pred)
19  
20      from sklearn.metrics import mean_squared_error
21  
22      return mean_squared_error(y_true, y_pred, squared=False)
23  
24  
25  def make_target_bins_for_reg_plots(
26      curr: pd.DataFrame, target_column, preds_column, ref: pd.DataFrame = None
27  ) -> pd.DataFrame:
28      df_for_bins = pd.DataFrame(
29          {
30              "data": "curr",
31              target_column: curr[target_column],
32              preds_column: curr[preds_column],
33          }
34      )
35      if ref is not None:
36          df_for_bins = pd.concat(
37              [
38                  df_for_bins,
39                  pd.DataFrame(
40                      {
41                          "data": "ref",
42                          target_column: ref[target_column],
43                          preds_column: ref[preds_column],
44                      }
45                  ),
46              ]
47          )
48      df_for_bins["target_binned"] = pd.cut(
49          df_for_bins[target_column], min(max(2, int(df_for_bins[target_column].nunique() / 3)), 10)
50      )
51      return df_for_bins
52  
53  
54  def fill_diagonal(data: pd.DataFrame, value):
55      """
56      Fill diagonal of DataFrame with given value in-place.
57      Args:
58          data: input DataFrame
59          value: value to fill with
60      """
61      min_el = min(data.shape)
62      for i in range(min_el):
63          data.iloc[i, i] = value