DOCSEARCHbyAlgolia
Submit Docs

modeltime Documentation

repository·master·Indexed 20 days ago

https://github.com/business-science/modeltime

A tidy time series forecasting framework for R that unifies classical statistical models, machine learning, and deep learning into a single workflow. It supports ARIMA, ETS, Prophet, and tidymodels/parsnip models, with an ecosystem including modeltime.h2o, modeltime.gluonts, modeltime.ensemble, modeltime.resample, and timetk.

Tokens
402
Snippets
1
Records
3
Agent score
20%

What's inside modeltime

  1. Overview of the modeltime forecasting capabilities

    master

    The modeltime framework unifies classical time series analysis and machine learning into a single workflow. It supports several modeling approaches:

    • Classical Forecast Models: Includes ARIMA and ETS (e.g., arima_reg(), arima_boost(), and exp_smoothing()).
    • Prophet: Integration with Facebook's Prophet algorithm via prophet_reg() and prophet_boost().
    • Tidymodels/Machine Learning: Any parsnip model can be used for forecasting, such as rand_forest(), boost_tree(), linear_reg(), mars(), and svm_rbf().

  2. Explore the modeltime ecosystem

    master

    The modeltime ecosystem consists of several specialized packages designed to work together for advanced forecasting tasks:

    • modeltime: Core machine learning and time series modeling.
    • modeltime.h2o: Integration with H2O for AutoML.
    • modeltime.gluonts: Deep learning capabilities using GluonTS.
    • modeltime.ensemble: Techniques for blending multiple forecasts.
    • modeltime.resample: Tools for backtesting and time series resampling.
    • timetk: Essential companion for feature engineering, data wrangling, and time series visualization.

  3. Install modeltime

    master

    You can install modeltime via CRAN for the stable version or via GitHub for the development version. Ensure you install dependencies to avoid issues.

    # CRAN version
install.packages("modeltime", dependencies = TRUE)

# Development version
remotes::install_github("business-science/modeltime", dependencies = TRUE)