Description
Orient a supplied metric change toward improvement.
Data workflows require explicit training scope, feature/label availability, missingness codes, group assignments and evaluation direction. Supplied training statistics are not proof that preprocessing avoided leakage. A local target-encoding term or retransformed prediction depends on assumptions that must be checked in the complete model and validation design.
Orient a supplied metric change toward improvement.
When to use this tool
Use Evaluation Metric Direction Sign when you already have signedMetricChange, higherIsBetter and need to orient a supplied metric change toward improvement.
How to use the tool and how it works
- Enter signedMetricChange, higherIsBetter. Use the units and conventions below.
- Run the tool. Correct any format, range, or cross-field input error before interpreting the output.
- Read improvementSignedChange. Use the output convention below.
Preserve signedMetricChange when higherIsBetter=true, otherwise negate it.1
| Input field | Meaning and units | Accepted input |
|---|---|---|
| signedMetricChange | Already supplied new-minus-old metric change; higherIsBetter chooses its improvement sign. | Required. Finite number. |
| higherIsBetter | Higher Is Better | Required. Boolean. |
| Output field | Interpretation | Returned shape |
|---|---|---|
| improvementSignedChange | Improvement Signed Change | Required. Finite number. |
{
"input": {
"signedMetricChange": 0.2,
"higherIsBetter": true
},
"output": {
"improvementSignedChange": 0.2
}
}Limitations and assumptions
- The change must already follow the intended new-minus-old convention. No metric, baseline, uncertainty, significance or model quality is calculated.
- 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.
- Numeric calculations use binary floating point unless exact decimal-integer arithmetic is explicitly stated. No hidden measurement tolerance or decimal rounding is added; nonfinite or out-of-contract results are rejected. Numeric signed-zero results are exported as zero.
Alternative or Complementary methods
Review the full data lineage, preprocessing fit, split design and evaluation procedure for model conclusions. Estimate required statistics only within the intended training folds and verify actual availability; local supplied-value transformations cannot establish those properties by themselves.
References
Similar or alternative tools
- metrics/brier-reference-class-contribution
Compare predicted and reference single-class squared probability errors.
- One Hot Reference Level Drop Flag Tool
Check whether an explicit one-hot category is the supplied reference level to drop without learning category order.