Reading a forecast

predict returns one forecast for each row you asked about. Each forecast contains:

Field	Meaning
timestamp	The date as a number of milliseconds. Use new Date(timestamp) to read it.
value	The forecast itself
trend	The long-term trend's share of the value
seasonalities	One entry per seasonal pattern, such as weekly or yearly
events	One entry per holiday or event
regressors	One entry per extra factor
additive	All additive pieces added up
multiplicative	All multiplicative pieces added up, as a fraction (0.1 means +10%)

Example

import { Effect } from "effect";
import { fit, predict, prophetFittingBackendLayer } from "effect-prophet";

const history = Array.from({ length: 90 }, (_, day) => ({
  timestamp: new Date(Date.UTC(2025, 0, 1 + day)).toISOString(),
  value: 100 + day * 0.5 + 10 * Math.sin((2 * Math.PI * day) / 7),
}));

const forecast = await Effect.runPromise(
  Effect.gen(function* () {
    const model = yield* fit(history);

    return yield* predict(model, ["2025-04-01T00:00:00.000Z"]);
  }).pipe(Effect.provide(prophetFittingBackendLayer)),
);

const [row] = forecast;

if (row !== undefined) {
  console.log("date:", new Date(row.timestamp).toISOString());
  console.log("forecast:", row.value.toFixed(2));
  console.log("trend:", row.trend.toFixed(2));

  for (const seasonality of row.seasonalities) {
    if (seasonality.mode === "additive") {
      console.log(`${seasonality.name}:`, seasonality.value.toFixed(2));
    }
  }
}

date: 2025-04-01T00:00:00.000Z
forecast: 137.18
trend: 145.00
weekly: -7.82

On this day the trend alone predicts 145. Tuesday is a quiet day of the week, so the weekly pattern takes away 7.82, leaving a forecast of 137.18.

Additive and multiplicative pieces

Most pieces are additive: they add or subtract a fixed amount, like "+20 sales on Saturdays". Their entries have a value.

A piece can also be multiplicative: it changes the value by a percentage, like "+15% on Saturdays". Their entries have a factor (0.15 means +15%) and a contribution, which is the same effect converted into your data's units. See Seasonality.

The pieces always add back up:

value = trend × (1 + multiplicative) + additive