Prophet
Prophet is a forecasting model developed by Facebook that is designed to make it easy for analysts and developers to produce high-quality forecasts for business time series data. It is based on a decomposable time series model with three main components: trend, seasonality, and holidays.
The trend component of the model is modeled using a piecewise linear or logistic curve, which allows the trend to change over time and to model non-linear trends. The seasonality component is modeled using Fourier series, which allows for multiple seasonalities (e.g. weekly, yearly) to be taken into account. The holidays component allows for the inclusion of special events that have a significant impact on the time series, such as Black Friday or New Year’s Day.
To fit the Prophet model, the user provides a time series dataset with timestamps and the corresponding values. Prophet then fits the model to the data, taking into account the trend, seasonality, and holidays components. The user can also specify additional regressors (predictors) to include in the model, such as weather data or promotional events.
Once the model is fit, the user can use it to make forecasts for a specified number of future periods. Prophet provides point forecasts as well as uncertainty intervals, which can be useful for decision-making.
