Description
Compare one multilabel prediction with actual labels as sets.
Multilabel predictions can be compared as literal sets or checked against a supplied hierarchy edge. Subset inclusion, exact set equality and distance through an asserted ancestor answer different questions; a local check need not verify label meaning or the complete hierarchy.
Compare one multilabel prediction with actual labels as sets.
When to use this tool
Use Multilabel Exact Match Single Observation when checking the stated label-set or parent/ancestor constraint.
How to use the tool and how it works
- Enter actual, predicted. Follow any array-alignment requirements below. Use the units and conventions below.
- Run the tool. Correct any format, range, or cross-field input error before interpreting the output.
- Read match, missing, extra. Inspect match before interpreting the associated values.
Compare one multilabel prediction with actual labels as sets. Deduplicate literal labels; return missing actual labels, extra predicted labels and match when neither exists.1
| Input field | Meaning and units | Accepted input |
|---|---|---|
| actual | Actual | Required. Array of text; maximum UTF-16 length 8192; maximum items 256. |
| predicted | Predicted | Required. Array of text; maximum UTF-16 length 8192; maximum items 256. |
| Output field | Interpretation | Returned shape |
|---|---|---|
| match | Match | Required. Boolean. |
| missing | Missing | Required. Array of text; maximum UTF-16 length 8192; maximum items 256. |
| extra | Extra | Required. Array of text; maximum UTF-16 length 8192; maximum items 256. |
{
"input": {
"actual": [
"a",
"b"
],
"predicted": [
"a",
"c"
]
},
"output": {
"match": false,
"missing": [
"b"
],
"extra": [
"c"
]
}
}Limitations and assumptions
- Duplicates and order do not affect equality; class meaning, partial credit and dataset-level accuracy are not inferred.
- Inputs must satisfy both the field types and the cross-field conditions above. Invalid input causes an error; a successful false flag or documented placeholder is a different result.
Alternative or Complementary methods
Validate the label vocabulary and complete hierarchy, and choose dataset-level set or hierarchical metrics appropriate to the prediction task.
References
Similar or alternative tools
- Multilabel Subset Violation Witness
Return required labels absent from a predicted set.
- Classifier Logit Tie Policy Tool
Compute classifier logit tie policy from explicit bounded inputs.