Intermediate · Data

Concept Drift

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. When the relationship between inputs and the target outcome shifts over time, even if inputs look the same.

Technical Definition

When the relationship between inputs and the target outcome shifts over time, even if inputs look the same.

How it works

Spam patterns evolve, fraudsters adapt, fashion changes — the same input today implies a different label than it did last year. Concept drift is harder to detect than data drift and usually requires labeled feedback from production to diagnose and fix.

Related Concepts

• Training — The process of adjusting a model's parameters so it learns patterns from labeled or unlabeled data.
• Data Drift — When the statistical distribution of inputs to a deployed model changes over time.
• Deployment — The process of moving a trained model into a production environment where it serves real users.