KeyError because of mismatch between outcome and positive label in data file

[KasperFyhn]

I get a KeyError in supervised_classification.py:77 which you seem to have anticipated.

In the block below, **report[pos_label] gives an error if your data file (as the one I got my hands on from you) contains 1's and 0's in the column corresponding to the outcome variable, e.g. political, instead of the actual positive label which the code seems to assume.

classifier.fit(X[train_index], y[train_index])
y_pred = classifier.predict(X[test_index])
report = classification_report(
    y[test_index], y_pred, output_dict=True
)
# The positive label is the same as the column name.
# But THIS MIGHT CHANGE so beware that then
# we have to relax the assumption in the next line.
pos_label = outcome
record = {
    "model": model,
    "outcome": outcome,
    "fold": i_fold,
    "accuracy": report["accuracy"],
    **report[pos_label],
}
records.append(record)

Looks like an easy fix.