Options

Time series forecasting

tonyboy9tonyboy9 Member Posts: 113 Contributor II
in Help

While RapidMiner uses the ARIMA operator, Facebook uses Prophet.

One key difference between ARIMA and Prophet is that the Prophet model accounts for “change points”, or specific shifts in trend in the time series. While it is technically possible to do this with ARIMA in R — it requires use of a separate package called AEDForecasting.

From Towards Data Science: 

https://towardsdatascience.com/arima-vs-prophet-forecasting-air-passenger-numbers-4e01b2d93608

Is RapidMiner currently working on making Facebook Prophet easier to use without all the "simple" Python code?



Best Answer

  • Options
    David_ADavid_A Administrator, Moderator, Employee, RMResearcher, Member Posts: 297 RM Research
    edited April 2022 Solution Accepted

    yes, we're not only working on it, but it's already available.
    If you have the latest version of Python Scripting extension installed, you will find a new operator called Python Forecaster.
    It allows you to build a Facebook Prophet (or other time series forecast model), that then can be used with Apply Forecast operator, like any other RapidMiner time series model.
    This gives you the best of both worlds, as you can use different Python libraries and the RapidMiner workflow design.

    Best,
    David




Answers

  • Options
    tonyboy9tonyboy9 Member Posts: 113 Contributor II
    Good day, David. I'm trying to duplicate your straightforward process for the time series prediction. The dataset I retrieved came from Samples, Time Series, Monthly Milk Production. Did I guess correctly? The Rename operator I chose came from Names and Roles. For "Rename the column...," how is that done in the Rename operator, assuming I've chosen the correct Rename? I will look at tutorials for Multiply, Python Forecaster and Apply Forecast. I am not a Python coder. Thanks for your time. Tony
  • Options
    [Deleted User][Deleted User] Posts: 0 Contributor I
    hi @tonyboy9 yes that is Monthly Milk Production data set. If you go to the Python Forecaster Help panel, you will see a tutorial for this process that David posted.


Sign In or Register to comment.