Selective Classification Accepted Count

Description

Calculate supplied nonabstained prediction count and conditional coverage.

Classification terms depend on label vocabularies, decision thresholds and denominator definitions. One class recall is not a full balanced-accuracy metric, and a row or column total is not automatically a precision or recall denominator. Empty or zero-total cases need an explicit defined flag or convention.

Calculate supplied nonabstained prediction count and conditional coverage.

When to use this tool

Use Selective Classification Accepted Count when you already have totalPredictions, abstentions and need to calculate supplied nonabstained prediction count and conditional coverage.

How to use the tool and how it works

Enter totalPredictions, abstentions. Use the units and conventions below.
Run the tool. Correct any format, range, or cross-field input error before interpreting the output.
Read acceptedPredictions, coverageFraction, defined. Inspect defined before interpreting the associated values.

Require nonnegative safe total and abstentions within total; accepted=total-abstentions, coverage=accepted/total or zero when total=0, with defined=total>0.^[1]

Input field	Meaning and units	Accepted input
totalPredictions	Total Predictions	Required. Safe integer; minimum 0.
abstentions	Abstentions	Required. Safe integer; minimum 0.

Output field	Interpretation	Returned shape
acceptedPredictions	Accepted Predictions	Required. Safe integer.
coverageFraction	Accepted/total, or zero placeholder when total=0.	Required. Finite number.
defined	False for zero total; the placeholder is not measured coverage.	Required. Boolean.

json

{
  "input": {
    "totalPredictions": 100,
    "abstentions": 20
  },
  "output": {
    "acceptedPredictions": 80,
    "coverageFraction": 0.8,
    "defined": true
  }
}

Limitations and assumptions

A zero-total fraction is a placeholder, not measured zero coverage. No prediction, correctness, risk, confidence threshold or appropriate abstention policy is evaluated.
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

Evaluate the complete prediction set, confusion structure and decision policy. Report missingness, class scope and uncertainty with the full metric; local threshold flags or one-cell fractions cannot characterize overall model performance.

References

Calculation and input contract: classification/selective-classification-accepted-count — tool.io

Similar or alternative tools

Binary Predicted Label Count Balance
Sum supplied binary confusion cells into predicted-label totals.
Balanced Accuracy Single Class Term
Calculate one class's recall and contribution to balanced accuracy.