Frank McQuillan created MADLIB-1261: ---------------------------------------
Summary: DT/RF- drop a categorical variable if it only has a single level Key: MADLIB-1261 URL: https://issues.apache.org/jira/browse/MADLIB-1261 Project: Apache MADlib Issue Type: Improvement Components: Module: Decision Tree, Module: Random Forest Reporter: Frank McQuillan Follow on from https://issues.apache.org/jira/browse/MADLIB-1258 In the JIRA 1258, for DT/RF we decided *not* to drop a categorical variable if it only has a single level. This avoided having to update `predict` (and other downstream functions) to handle the varying cat features across groups. However this is not the most efficient implementation. The alternative is drop a categorical variable if it only has a single level, and to make the necessary downstream changes including to `predict` . -- This message was sent by Atlassian JIRA (v7.6.3#76005)