RapidMiner's New Time Series Extension
I’m really liking the overhaul of the old Value Series extension. RapidMiner has been building a new Time Series extension that started with the great ARIMA operator.
Now they’re adding new Windowing and Sliding Window operators. My notes below the video.
- New time series datasets added to updated Time Series extension (gas station prices)
- Plus three more sample data templates
- New Windowing operator, easier to use parameters.
- New Indices parameter / New Horizon Width and Horizon Offset
- New Process Windows operator. It’s like a Loop for Time Series data - NICE!
- New Forecast Validation. Appears to be a redo the old Sliding Window Validation operator
- In the Testing side of the Forecast Validation operator, you don’t need to use Apply Model
- Difference between Forecast Validation (FV) and Cross Validation (CV) operator is that the model delivered by FV is always the LAST model that was trained, not like CV that trains over the entire data in the final iteration