Member Junction
    Preparing search index...

    Interface MLInferenceResultPayload

    The structured payload a successful scoring run carries on its RecordResult.

    interface MLInferenceResultPayload {
        class?: string;
        drivers?: { feature: string; value: number }[];
        modelId: string;
        problemType: ProblemType;
        score: number;
        scoredAt: string;
        target: string;
    }
    Index

    Properties

    class? drivers? modelId problemType score scoredAt target

    Properties

    class?: string

    Predicted class label, present for classification problems.

    drivers?: { feature: string; value: number }[]

    Top signed per-record drivers behind this prediction (P1-5), present only for models with exact per-row attribution (linear). Feature names are post-preprocessing output columns; UIs collapse/humanize them. value > 0 pushed the score up, < 0 down. Absent for tree/ensemble/multiclass models.

    modelId: string

    The model that produced the prediction (lineage).

    problemType: ProblemType

    Classification or regression.

    score: number

    Numeric model output: probability (classification) or value (regression).

    scoredAt: string

    ISO-8601 UTC timestamp of when this prediction was produced. Lets write-back persist a dedicated "last scored at" column on the target entity via OutputMapping ({ "LastScoredAt": "$.scoredAt" }) — distinct from the row's whole-record __mj_UpdatedAt, which moves on any edit, not just scoring.

    target: string

    The target/label the model predicts.

    Settings

    Member Visibility

    On This Page

    Properties