P65Machine LearningAdvancedDiagnose14 min

Diagnose Unstable Feature Importance Across Model Runs

Explain shifting importance among correlated predictors.

#diagnose#feature importance#correlation

Scenario

tenure_months, account_age_days, and contract_start_date swap tree-importance ranks across seeds while their combined importance stays 0.42.

Problem statement

Diagnose the instability.

Your Task

  1. Explain rank changes.
  2. State defensible conclusions.
  3. Avoid causal claims.

Related Concepts

CorrelationCausality

Ready to count this problem?

Mark it complete when you have reasoned through the solution in your own words.