Interface DataPoint

prediction datapoint structure

complete prediction result with rankings, scores, and propensities. contains all information needed for feedback and analysis.

interface DataPoint {
    actions: string[];
    context_features: Record<string, string>;
    action_features: Record<string, string>[];
    rewards: any;
    ordering: {
        [k: string]: number;
    };
    scores: {
        [k: string]: number;
    };
    is_randomized: boolean;
    propensities: {
        [k: string]: number;
    };
}

Properties

actions context_features action_features rewards ordering scores is_randomized propensities

Properties

actions

actions: string[]

array of action identifiers that were ranked

context_features

context_features: Record<string, string>

context features provided in request

action_features

action_features: Record<string, string>[]

action features for each ranked item

rewards

rewards: any

reward values (if applicable)

ordering

ordering: {
    [k: string]: number;
}

item rankings (item_id -> rank position)

scores

scores: {
    [k: string]: number;
}

prediction scores for each item (item_id -> score)

is_randomized

is_randomized: boolean

whether prediction used exploration/randomization

propensities

propensities: {
    [k: string]: number;
}

propensity scores for each item (item_id -> propensity)

Settings

Member Visibility

On This Page

Properties