Struct ForecastEngine 

pub struct ForecastEngine { /* private fields */ }

Forecast engine — produces predictions from historical data.

Uses a hybrid approach:

1. Linear extrapolation from the slope of recent samples
2. EWMA smoothing to reduce noise sensitivity
3. Confidence from residual variance (how well linear fit matches data)

Implementations

impl ForecastEngine

pub const fn new() -> Self

Create a new forecast engine with default alpha (0.3).

pub const fn with_alpha(alpha: f32) -> Self

Create with a custom EWMA alpha.

pub fn forecast( &self, history: &VecDeque<MetricSample>, horizon: usize, ) -> Forecast

Forecast from a history of samples.

Combines linear extrapolation (for trend) with EWMA (for noise resistance). Confidence is derived from how well the linear model fits the data (R²-like measure from residual variance).

Outlier values are clamped to ±3 standard deviations from the mean before fitting, preventing a single extreme spike from dominating the forecast.

Trait Implementations

impl Default for ForecastEngine

fn default() -> Self

Returns the “default value” for a type.