@logistics-ts/forecasting

Demand forecasting for logistics-ts: moving average, exponential smoothing (SES, Holt ±damped, Holt-Winters), intermittent-demand methods (Croston/SBA/TSB), seasonal decomposition, rolling-origin backtesting, accuracy metrics, and autoForecast, which classifies the series and picks the lowest-MASE method for you. Every method returns an Explained<T> Forecast.

Install

npm i @logistics-ts/forecasting

What's in it

• autoForecast — classifies the demand pattern (smooth / erratic / intermittent / lumpy), backtests the candidate methods for that quadrant, and returns the lowest-MASE one. You don't pick the method.
• Point methods: movingAverage, ses, holt (±damped), holtWinters (additive/multiplicative) for smooth/erratic demand; croston, sba, tsb for intermittent/lumpy demand.
• seasonalDecompose — classical additive/multiplicative decomposition.
• backtest — rolling-origin backtesting for any Forecaster.
• Metrics: mae, rmse, mape, smape, mase, bias — prefer mase for intermittent series (mape is undefined at zero demand).

Quick start

import { bucketize, generateExampleData } from '@logistics-ts/core'
import { autoForecast } from '@logistics-ts/forecasting'

const { demand } = generateExampleData({ items: 1, periods: 24, seed: 3 })
const series = bucketize(demand, 'month')[0]
const quantities = series.buckets.map((b) => b.quantity)

const f = autoForecast(quantities, { horizon: 3 })
console.log(f.value.forecast) // number[] — next 3 periods
console.log(f.method)         // e.g. 'auto-holt' — the winning method
console.log(f.reasoning)      // why it was chosen (pattern, candidates, MASE scores)

Feed bucketize's dense, zero-filled output — not a compacted list of nonzero-only sales — or the intermittent-demand statistics (ADI, CV²) and exponential-smoothing recursions will be wrong.

In the umbrella package

@logistics-ts/forecasting is re-exported as the forecasting namespace from logistics-ts. It depends on @logistics-ts/core and @logistics-ts/classification (autoForecast routes by demand-pattern classification).

Links

• Full docs, API map, and examples
• Other logistics-ts packages

License

MIT