Forecast Horizon

This component visualizes how different error metrics change as a forecast horizon grows. Use this component to better understand how far into the future your forecasting model is reliable.

A line plot displays the error metrics changing over time.

The time on the x-axis is measured in forecast length. Each unit is equal to one record into the future. In fact this component only works on an equally spaced Time Series. For example, if input data is hourly, the forecast length 24 represents one day, 24 hours. If the input data is every two hours it would represent two days, 48 hours. Keep this in mind when interpreting the chart.

Metrics on the y-axis displayed in the line plot are cumulative, while metrics displayed in the adjacent tile view are based on the entire input table. All metrics are exported at the output of the component by forecast length.

Options

Smooth

Check this box to smooth calculated metrics with a moving average calculation.

Reference Column

Select the column containing true values.

Predicted Column

Select the column with the predicted values.

Metrics to Display

Select which metrics to display in the line plot. All metrics will still be output at the end of the component.

Nodes

• Rule Engine Variable8 ×
• Table Row to Variable5 ×
• Column Filter4 ×
• Row Filter3 ×
• CASE Switch Variable (End)2 ×
• Show all 29 nodes

Extensions

• FeatureKNIME Base nodes
• FeatureKNIME JavaScript Views
• FeatureKNIME JavaScript Views (Labs)
• FeatureKNIME Javasnippet
• FeatureKNIME Math Expression (JEP)
• Show all 7 modules

Links

• No links available