Skip to content

R/Shiny app showcasing forecasting methods for multiple time series

Notifications You must be signed in to change notification settings

rafabelokurows/shiny_time_series_forecasting

Repository files navigation

Multiple Time Series Forecasting Shiny App

Shiny App that highlights a few findings I made when forecasting multiple time series with modeltime and tidymodels, two great packages for Time Series forecasting in R. Data came from the timetk package and hopefully, this app and framework can be applied to other time series datasets.
Live demo Time Series Multiple Forecasting.

Table of Contents

General Information

Training several methods on multiple time series at once using R

Technologies Used

  • R
  • modeltime
  • tidymodels
  • Shiny

Features

List the ready features here:

  • How to apply several ML and traditional time series forecasting methods (Random Forest, SVM, ARIMA, XGBoost, Prophet, and others)
  • Showcase a quick way of communicating/evaluating results for each individual time series
  • Aggregate forecasts to simulate a retail company forecasting sales for many departmens/groups.

Usage

  • .rds files contain data
  • .Rmd contains code for Shiny app
  1. Place .rds and .Rmd files within the same folder
  2. Run .Rmd once you have all packages installed.

Project Status

no longer being worked on: on the to-do list for improvement

Room for Improvement

  • Adding some more time series
  • Improving Feature Engineering
  • Adding Deep Learning algorithms

Acknowledgements

Thanks to all developers of the packages used in this project.

Contact

Created by @rafabelokurows - feel free to contact me!

About

R/Shiny app showcasing forecasting methods for multiple time series

Topics

Resources

Stars

Watchers

Forks