facebook/prophet
Prophet
Prophet is a time series forecasting library and decomposition tool that uses an additive regression model to predict future values. It functions as an uncertainty estimation tool, calculating confidence intervals and error metrics to quantify the risk associated with future predictions.
The project is distinguished by its ability to incorporate human-interpretable parameters for model tuning and its use of Bayesian inference for parameter estimation. It supports the integration of external regressors and special event modeling to account for the impact of holidays and specific dates on forecasted trends.
The system provides a comprehensive suite for time series analysis, including trend changepoint detection, saturating growth constraints, and seasonal effect modeling using Fourier series. It includes capabilities for handling irregular data, validating performance via cross-validation, and visualizing historical fits and seasonal components.
Features
- Time Series Forecasting - Provides a comprehensive library for predicting future numeric values using additive time series models.
- Additive Models - Implements an additive regression model that combines non-linear trends with multiple seasonal patterns.
- Additive Regression Models - Implements a forecasting approach that combines non-linear growth, seasonal patterns, and holiday effects.
- Piecewise Linear Trends - Fits piecewise-linear trends by detecting automatic or manual changepoints to handle growth rate shifts.
- Uncertainty Estimation - Calculates confidence intervals and predictive variance to quantify the uncertainty and risk of future predictions.
- Seasonal Pattern Modeling - Implements an additive model that combines trend, seasonal effects, and holidays to predict future values.
- Time Series Decomposition - Provides a framework for separating time series data into its underlying trend, seasonal, and residual components.
- Fourier-Series Seasonality - Models periodic daily, weekly, and yearly seasonal patterns using Fourier series of sine and cosine waves.
- Saturating Growth Models - Constrains predicted trends using carrying capacity and floors to model realistic growth boundaries.
- Interpretable Parameter Tuning - Allows adjusting predictions via human-interpretable parameters to incorporate domain-specific expertise into the model.
- Forecast Accuracy Validation - Provides historical cross-validation and error metrics to validate the accuracy of time-series forecasts.
- Prediction Visualization - Generates plots that visualize historical fits, predicted trends, and decomposed seasonal components.
- Predictive Trend Modeling - Models predictive trends that are robust to outliers and abrupt shifts in the time series.
- Bayesian Inference - Employs Stan for Bayesian inference and Hamiltonian Monte Carlo sampling to estimate model parameters.
- Irregular Time Series - Handles irregular time series data, including missing values and outliers, without requiring extensive manual cleaning.
- Demand Planning - Supports business demand planning by accounting for holidays and carrying capacity in growth forecasts.
- Saturating Growth Models - Allows constraining predicted growth using a carrying capacity to model realistic upper or lower boundaries.
- External Regressors - Integrates additional independent time-varying variables as predictors to enhance the accuracy of forecasts.
- Special Event Modeling - Models the specific impact of holidays and special dates on forecasted trends and seasonality.
- Changepoint Detection - Detects and adjusts for shifts in growth rates to improve the accuracy of long-term projections.
- Time Series - Automatic procedure for forecasting time series data.
- Time Series Analysis - Tool for forecasting time series data with seasonal effects.
- Time Series Analysis - Tool for forecasting time series data with seasonal effects.
Star history
How this analysis was created: This summary and feature list are AI-generated from collected project material and can contain mistakes. Stars, license and language are imported from GitHub. Inclusion does not mean that we have tested or audited this project. Check the source documentation for any feature you depend on. Learn more on our About page.
