The Prophet Model
Prophet is a forecasting approach that models a series as a sum of interpretable components. It targets business data with strong seasonality and known special days.
The building blocks
Prophet writes the forecast as an additive combination of pieces.
- Trend: a piecewise curve that can bend at automatically chosen change points.
- Seasonality: smooth repeating patterns, often modeled with a Fourier series for weekly and yearly cycles.
- Holidays: named dates that get their own effect, such as a sales spike on a public holiday.
Why teams like it
- It is robust to missing data and outliers.
- Analysts can add domain knowledge by listing holidays and adjusting how flexible the trend is.
- It produces uncertainty intervals along with the point forecast.
Trade offs
Prophet shines on human scale calendar data but is not designed for high frequency signals or short noisy series. Because the trend is piecewise, very flexible settings can overfit the recent past and project unrealistic future bends.
Key idea
Prophet sums a flexible trend, Fourier seasonality, and holiday effects into a readable, robust forecaster for calendar driven business data.