I am currently using pandas.crosstab to generate the confusion matrix of my classifiers after testing. Unfortunately, sometimes my classifier fails, and classifies every signal as a single label (instead of multiple labels). pandas.crosstab generates a single vector (or a non-square matrix) in that case instead of a square matrix.
As example, my ground truth would be
true_data = pandas.Series([1, 1, 2, 2, 3, 3, 4, 4, 5, 5])
and my predicted data is
pred_data = pandas.Series([3, 3, 2, 3, 2, 1, 1, 3, 4, 1])
Applying pandas.crosstab(true_data, pred_data, dropna=False) gives
col_0 1 2 3 4
row_0
1 0 0 2 0
2 0 1 1 0
3 1 1 0 0
4 1 0 1 0
5 1 0 0 1
Is there a way to get
col_0 1 2 3 4 5
row_0
1 0 0 2 0 0
2 0 1 1 0 0
3 1 1 0 0 0
4 1 0 1 0 0
5 1 0 0 1 0
instead, i.e. leaving the matrix square and filling the missing labels with 0?
You could create a zeros array of the required shape and then replace a portion of the array with the crosstab
xtab = pd.crosstab(pred_data, true_data, dropna=False).sort_index(axis=0).sort_index(axis=1)
all_unique_values = sorted(set(true_data) | set(pred_data))
z = np.zeros((len(all_unique_values), len(all_unique_values)))
rows, cols = xtab.shape
z[:rows, :cols] = xtab
square_xtab = pd.DataFrame(z, columns=all_unique_values, index=all_unique_values)
Output
1 2 3 4 5
1 0.0 0.0 1.0 1.0 1.0
2 0.0 1.0 1.0 0.0 0.0
3 2.0 1.0 0.0 1.0 0.0
4 0.0 0.0 0.0 0.0 1.0
5 0.0 0.0 0.0 0.0 0.0
I haven't thought / tested yet if this approach will work if the mismatch is in the "middle" - as in, if, e.g., pred_data = [1, 2, 4, 5] and true_data = [1, 2, 3, 4]