This extension provides two new operators, Foreacast (Univariate) and Forecast (Multivariate) which allow simple but powerful forecasting of time series.

=== Version 0.1.6 ===

Fixed a bug in depenencies which made 0.1.5 unable to load in normal studios.

=== Version 0.1.5 ===
Added a new operator 'Visualize Forecast Results' which allows you to visualize your results as pngs on your disc. this needs the new animated plots extension

=== Version 0.1.4 ===
Holt-Winters is now also optimizing its seasonality. This makes the operator slower, but better

=== Version 0.1.3 ===
Moved to 5-fold sliding window val for Univariate
Fixed a bug that ARIMA did not use an heuristic but constants for p/q
Added errors if attributes do not exist
Univariate is now also working if your time series attribute has a special role (it is ignored).


=== Version 0.1.2 ===

Both operators now require 48 examples to work.

Linear Model is now using a Lambda optimization with X-Val internally.

Multivariate is not using the last value of each attribute anymore Instead Multivariate uses the last k values of the smoothed (moving avg) time series'. k is defined as a value of window length but max 5.

Product Details

Version 0.1.6
File size 47 kB
Downloads 6424 (10 Today)6424 downloads
Vendor RapidMiner Labs
Category Machine Learning
Released 9/1/20
Last Update 9/1/20 3:58 PM
License AGPL
Product web site
Rating 0.0 stars(0)