Explaining Concept Shift with Interpretable Feature Attribution

TL;DR

This paper introduces SGShift, a method for detecting and explaining concept shift in tabular data by identifying features responsible for performance drops, using a GAM-based approach with extensions for false discovery control.

Contribution

The paper proposes SGShift, a novel GAM-based model that detects and attributes concept shift to specific features, improving interpretability and accuracy over baseline methods.

Findings

SGShift achieves AUC > 0.9 in identifying shifted features.

Recall exceeds 90% in various experiments.

Outperforms baseline methods by 2-3 times in feature detection.

Abstract

Regardless the amount of data a machine learning (ML) model is trained on, there will inevitably be data that differs from their training set, lowering model performance. Concept shift occurs when the distribution of labels conditioned on the features changes, making even a well-tuned ML model to have learned a fundamentally incorrect representation. Identifying these shifted features provides unique insight into how one dataset differs from another, considering the difference may be across a scientifically relevant dimension, such as time, disease status, population, etc. In this paper, we propose SGShift, a model for detecting concept shift in tabular data and attributing reduced model performance to a sparse set of shifted features. SGShift models concept shift with a Generalized Additive Model (GAM) and performs subsequent feature selection to identify shifted features. We propose further extensions of SGShift by incorporating knockoffs to control false discoveries and an absorption term to account for models with poor fit to the data. We conduct extensive experiments in synthetic and real data across various ML models and find SGShift can identify shifted features with AUC $>0.9$ and recall $>90\%$, often 2 or 3 times as high as baseline methods.