Struct MultivariateForecastResult 

pub struct MultivariateForecastResult {
    pub timestamps: Vec<i64>,
    pub series: Vec<String>,
    pub predictions: Vec<Vec<f64>>,
    pub target: String,
    pub predictor_importance: Vec<(String, f64)>,
}

Result of a multivariate forecast.

Fields

timestamps: Vec<i64>

Forecast timestamps (nanoseconds since epoch).

series: Vec<String>

The series each row of predictions belongs to, in the same order.

This exists because the two implementors of MultivariateForecastModel return different numbers of rows for the same call, and nothing used to say so. MultiLinearRegression is many-predictors-to-one-target and returns one row; VarModel fits every equation of the system and returns one row per series. A caller reaching for predictions[0] therefore got the requested series from one and series index 0 from the other — a plausible number belonging to a different series, with nothing in the type to catch it.

predictions: Vec<Vec<f64>>

Predicted values, one row per entry of series.

target: String

The series named by target_idx at fit.

Prefer target_forecast to indexing predictions directly.

predictor_importance: Vec<(String, f64)>

Per-predictor importance / coefficients.

Implementations

impl MultivariateForecastResult

pub fn for_series(&self, name: &str) -> Option<&[f64]>

The forecast for one named series, if this model produced it.

pub fn target_forecast(&self) -> Option<&[f64]>

The forecast for the series the caller asked fit for.

Correct for every implementor, which is the point: it is what predictions[0] was reached for and only sometimes was.

Trait Implementations

impl Clone for MultivariateForecastResult

fn clone(&self) -> MultivariateForecastResult

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for MultivariateForecastResult

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.