A feature's importance to predictions shifts over time. What is this?
Why D is the answer
Feature attribution drift
Feature attribution drift means a feature's importance to predictions has shifted over time.
In detail
- Importance moving over time is feature attribution drift.
- It can warn before accuracy falls.
Use case
An illustrative scenario, not a real incident.
Price mattered most at launch, but now location drives predictions. That is attribution drift.
See it in different ways
flowchart TD A["Model issue"] --> B["Skew: training and serving data differ"] A --> C["Data drift: inputs change over time"] A --> D["Concept drift: the relationship changes"] A --> E["Attribution drift: feature importance shifts"]
mindmap
root((Decay))
Data drift
Inputs changed
Concept drift
Relationship changed
Skew
Training differs from serving
Attribution drift
Feature importance movedstateDiagram-v2 state "Price leads" as s0 state "Location leads" as s1 state "Investigate" as s2 [*] --> s0 s0 --> s1: behavior changes s1 --> s2 s2 --> [*]
Key point: Attribution drift tracks feature importance.
Why the other options are wrong
- A. Hyperparameter tuning
- Tuning searches for values.
- B. Data parallelism
- That is about training.
- C. A canary rollout
- That is about rollout.