ClassificationQualityByClassRenderer

[sara-vallejo]

I am using the ClassificationQualityByClass metric for my classificaiton ML model. In the reference data I have 7 labels, however, in the current data there are only example for 3 of them. When I run a report with this data and metric, the tables for current and reference only show the metrics for the 3 labels that appear on the "current" data, but not for all 7 labels in the reference data.

I checked the renderer class for this metric, and I can see that for the reference plot, the names being used for x-axis are the ones from the current_metrics frame, instead of the reference_matrix_frame. Is this done on purpose? If not, I think a quick fix that would return the metrics for all labels in the reference data is using

ref_names = ref_metrics_frame.columns.tolist()
x = list(map(str,ref_names))

for the reference plot (line 169 in evidently/metrics/classification_performance/quality_by_class_metrics.py).

Otherwise, is there any way to overwrite original renderer plots?