<br />

Unified decision output for a single ClassifierQuestion.

**Signature:**

    export declare interface ClassifierDecision 

## Properties

| Property | Modifiers | Type | Description |
|---|---|---|---|
| [id](https://developers.google.com/edge/api/mediapipe/js/tasks-decision.classifierdecision#classifierdecisionid) |   | string | Question identifier matching `ClassifierQuestion.id`. |
| [label](https://developers.google.com/edge/api/mediapipe/js/tasks-decision.classifierdecision#classifierdecisionlabel) |   | string | Winning option label (`"true"`/`"false"` for binary, winning label for categorical/ordinal). |
| [confidence](https://developers.google.com/edge/api/mediapipe/js/tasks-decision.classifierdecision#classifierdecisionconfidence) |   | number | Normalized certainty confidence in \[0.0, 1.0\]. |
| [probability](https://developers.google.com/edge/api/mediapipe/js/tasks-decision.classifierdecision#classifierdecisionprobability) |   | number | Binary only: calibrated P(true) in \[0.0, 1.0\]. |
| [expectedScore](https://developers.google.com/edge/api/mediapipe/js/tasks-decision.classifierdecision#classifierdecisionexpectedscore) |   | number | Ordinal only: continuous expected score across levels, sum(level_i \* p_i). |
| [probabilities](https://developers.google.com/edge/api/mediapipe/js/tasks-decision.classifierdecision#classifierdecisionprobabilities) |   | ClassifierOptionProbability\[\] | Full calibrated probability distribution over candidate options. |
| [predictionSet](https://developers.google.com/edge/api/mediapipe/js/tasks-decision.classifierdecision#classifierdecisionpredictionset) |   | string\[\] | Optional conformal prediction set covering \>= 90% probability mass. |

## ClassifierDecision.id {#classifierdecisionid}

Question identifier matching `ClassifierQuestion.id`.

**Signature:**

    id: string;

## ClassifierDecision.label {#classifierdecisionlabel}

Winning option label (`"true"`/`"false"` for binary, winning label for categorical/ordinal).

**Signature:**

    label: string;

## ClassifierDecision.confidence {#classifierdecisionconfidence}

Normalized certainty confidence in \[0.0, 1.0\].

**Signature:**

    confidence: number;

## ClassifierDecision.probability {#classifierdecisionprobability}

Binary only: calibrated P(true) in \[0.0, 1.0\].

**Signature:**

    probability?: number;

## ClassifierDecision.expectedScore {#classifierdecisionexpectedscore}

Ordinal only: continuous expected score across levels, sum(level_i \* p_i).

**Signature:**

    expectedScore?: number;

## ClassifierDecision.probabilities {#classifierdecisionprobabilities}

Full calibrated probability distribution over candidate options.

**Signature:**

    probabilities: ClassifierOptionProbability[];

## ClassifierDecision.predictionSet {#classifierdecisionpredictionset}

Optional conformal prediction set covering \>= 90% probability mass.

**Signature:**

    predictionSet?: string[];