Review causal correlations
After completion of the forecast run, use the Causal Analysis page to understand what causals (internal causals, Trend Cloud causals, and derived causals) are most affecting the forecast. You can review the key error metric, understand the models used (stat + ML models), and understand the parameters optimized. You can analyze the results by product and location.
Before you review your forecasts, be sure you have run the appropriate scenarios.
- Log into the Demand Modeler app using the link provided by your administrator.
- Select the Output Overview
button.
- Select the Causal Analysis
button. The Causal Analysis page appears.
- At the top of the page, select a scenario to use from the Scenario field.
- Select the Filters button to select the slice of the forecast to analyze. The visualizations are updated according to your selections.
- Select the Detailed Correlation, Feature Type, or Variable Importance tab, as needed.
- Review the results.
Causal Analysis tab
Visualization | Description |
---|---|
Detailed Correlation tab | |
Correlation with Demand graph | This graph displays the correlation of causals with demand. You can hover over the scatter plot to see detailed correlation information. |
bar chart | This chart displays the number of causals by correlation thresholds. Clicking on a column filters the table below. |
table | This table displays details of the causals that were selected by the prediction engine. It can be filtered by clicking one of the bars in the bar graph above it, or using the Filters button. |
Feature Type tab | |
User Defined Feature chart | This graph displays the number of causals by user defined causals. |
Trend Cloud Feature chart | This graph displays the number of causals by Trend Cloud causals. |
Variable Importance tab | |
Causal Influence chart |
This chart displays the influence of the various causals on the forecast. If you do not filter down to the individual LOD, the top 5 causals are aggregated at the LOD aggregation you have chosen. For example, if you choose to filter down to a value in Dimension 1, you are able to see the top 5 causals for Dimension 1. |
table | This table displays the influence, absolute value, and normalized value of the causals shown in the chart. The table is filtered by selections made in the chart. |
Last modified: Friday May 12, 2023