ValueError from `eval_model` function within Chapter 4: 02-Baseline Forecasts using darts.ipynb Cell 12

[beliu]

When running the eval_model function within Chapter 4: 02-Baseline Forecasts using darts.ipynb at cell 12, I get the following error:

Cell In[12], line 4
      2 naive_model = NaiveSeasonal(K=1)
      3 with LogTime() as timer:
----> 4     y_pred, metrics = eval_model(naive_model, ts_train, ts_val, name=name)
      5 metrics['Time Elapsed'] = timer.elapsed
      6 metric_record.append(metrics)

Cell In[10], line 11
      4 model.fit(ts_train)
      5 y_pred = model.predict(len(ts_test))
      6 return y_pred, {
      7     "Algorithm": name,
      8     "MAE": mae(actual_series = ts_test, pred_series = y_pred),
      9     "MSE": mse(actual_series = ts_test, pred_series = y_pred),
     10     "MASE": mase(actual_series = ts_test, pred_series = y_pred, insample=ts_train),
---> 11     "Forecast Bias": forecast_bias(actual_series = ts_test, pred_series = y_pred)
     12 }

File ~/Modern-Time-Series-Forecasting-with-Python/src/utils/ts_utils.py:102, in forecast_bias(actual_series, pred_series, intersect, reduction, inter_reduction, n_jobs, verbose)
    100 else:
    101     y_true, y_pred = _get_values_or_raise(actual_series, pred_series, intersect)
--> 102 y_true, y_pred = _remove_nan_union(y_true, y_pred)
    103 y_true_sum, y_pred_sum = np.sum(y_true), np.sum(y_pred)
...
   5348                          'length of {}'.format(N))
   5350     # optimization, the other branch is slower
   5351     keep = ~obj

ValueError: boolean array argument obj to delete must be one dimensional and match the axis length of 1488

I went into ts_utils.py and made the following changes:

def _remove_nan_union(array_a: np.ndarray,
                      array_b: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
    """
    Returns the two inputs arrays where all elements are deleted that have an index that corresponds to
    a NaN value in either of the two input arrays.
    """

    isnan_mask = np.logical_or(np.isnan(array_a), np.isnan(array_b))

    # I added the line below
    isnan_mask = isnan_mask.reshape(-1,)

    return np.delete(array_a, isnan_mask), np.delete(array_b, isnan_mask)

This resolved the ValueError but I do not know if there are any side-effects of my modification.

[abdullahozkoc]

Thank you for the modification. I added your change to my code, and when I generated the graph afterward, I found that it matched the one in the book. I think your change didn't affect the result.

[Laoban-man]

That modification corrected the problem for me too, thanks.