DataTerminology.comNEW
Also known as: Facebook Prophet
A decomposable forecasting library modeling trend, seasonality, and holidays with strong defaults for business time series.
DESCRIPTION
Prophet is a decomposable time series model and library designed for robust business forecasting. It models trend (piecewise linear or logistic), multiple seasonalities, and holiday effects. Strong defaults and automated changepoint detection enable quick, reliable baselines with minimal tuning. Implementations in Python and R compile through Stan for Bayesian inference. Cross-validation utilities assist in performance assessment and parameter tuning. It is well-suited to messy data with missing values and outliers.
ALTERNATIVE DEFINITIONS
Prophet Documentation
Prophet is a procedure for forecasting time series data based on an additive model where non-linear trends are fit with yearly, weekly, and daily seasonality, plus holiday effects.
Highlights additive decomposition with multiple seasonalities and holidays.
Meta Research
Prophet is a decomposable time series model with three main components: trend, seasonality, and holidays.
Condenses Prophet’s core three-component decomposition.
RELATED TERMS
Temporal Data Holt-Winters Fourier Features Stan Holidays Additive Model NeuralProphet Cross-Validation
COMPETING TERMS
ARIMA Exponential Moving Average
KNOWLEDGE AREAS
Data Science AI MLOps Platform & Tooling
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer Analytics Engineer
EXTERNAL SOURCES
- Wikipedia - Prophet (software)
- Prophet Documentation - Quick start and examples
- CRAN - R package reference
TOP QUESTIONS
When should I use Prophet instead of ARIMA?
Use Prophet for business series with multiple seasonalities, holidays, and changepoints where defaults matter. ARIMA may excel with strong autocorrelation structures and careful identification.
Category: Comparison
How does Prophet model changepoints?
It uses sparse priors over potential changepoints to allow piecewise linear or logistic trends, adapting slope while preventing overfitting.
Category: Concept
Does Prophet handle missing data and outliers?
Yes. Prophet is robust to missing values and outliers via trend modeling and scale parameters, often requiring minimal preprocessing.
Category: Capabilities
Can I add custom seasonality in Prophet?
Yes. You can specify additional seasonalities with chosen periods and Fourier orders to model domain-specific cycles.
Category: How-to
What are Prophet’s limitations?
It focuses on additive structures; complex autocorrelation may be under-modeled. It is univariate and less suited for intricate exogenous interactions without careful feature engineering.
Category: Limitations
How do I evaluate Prophet forecasts?
Use cross-validation utilities, metrics like MAPE and RMSE, and backtesting with rolling-origin splits to assess generalization.
Category: How-to
Does Prophet support multiplicative seasonality?
Yes. You can set seasonality_mode to “multiplicative” to scale seasonal effects by the level of the series.
Category: Configuration
What holidays can Prophet include?
Built-in country-specific holidays exist; you can add custom events, promotions, and special days as regressors.
Category: Features
How does Prophet perform with short time series?
It can work, but limited data reduces reliable changepoint and seasonality estimation. Incorporating domain priors helps.
Category: Risks