[SPARK-14216] ML tree models should have a standardized, reusable feature importance test

• Type: Test
• Status: Resolved
• Priority: Minor
• Resolution: Incomplete
• Component/s: ML, Tests
• Labels: bulk-closed
• Assignee: Unassigned
• Reporter: Seth Hendrickson
• Created: 28/Mar/16 22:02
• Updated: 21/May/19 04:14
• Resolved: 21/May/19 04:14

Description

ML models that have a featureImportances attribute all need to perform essentially the same test in their respective testing suites. Currently a fairly primitive test is basically copy/pasted in each test suite. It would be nice to create a single testing method that can be reused for all ML models with feature importance. It would also be nice to increase the robustness of the current test.

Issue Links

• Is contained by: SPARK-14045 DecisionTree improvement umbrella