Feature Mapping

The process of mapping Trend Cloud features to LODs consists of two steps:

  1. Modelers map feature_names to feature_groups through the feature_groups table.

  2. Modelers map those feature_groups to LODs through the feature_assignment table. One strategy may be to create a unique feature_group for every LOD, especially if the number of LODs is small. Another strategy may be to create one feature_group per category of LOD, where categories could be the values of a particular dimension or level across the LODs.

Modelers who prefer to map all features to all LODs do not need to create feature_assignment mappings for individual LODs, nor is the feature_groups table needed at all. Instead, they can simply add one row to the feature_assignment table, in which dimension_1=”ALL”, dimension_2=”ALL”, dimension_3=”ALL”, feature_group=”ALL”, and assignment_type=True. Alternatively, to map all features to all LODs, but with a few exceptions, the modeler should pass this row along with a set of LOD-to-feature group mappings in which assignment_type=False, thereby indicating that those LODs will not be mapped to those feature groups.

Similarly, to map certain feature_groups to all LODs with a particular value in one dimension (or a particular combination of values in two dimensions), modelers can use the “ALL” value in the other dimensions for the same effect. For example, if dimension_1=”CA”, dimension_2=”ALL”, dimension=”ALL”, and feature_group=feature_group_CA, all LODs with dimension_1=”CA” are mapped to feature_group_CA.

A final option is to further subset LOD-to-feature mappings through the Correlation Calculator; therefore, it is reasonable for modelers to create their mappings on a theoretical basis at this stage and let the Correlation Calculator subset them further based on empirical correlations later.

Modelers should create a copy of their scenario before using this functionality to retain a copy of their original mappings in the Demand Modeler app, even after Correlation Calculator overwrites them in the other scenario.

Last modified: Friday May 12, 2023

Is this useful?