Explaining Concept Shift with Interpretable Feature Attribution
This addresses the issue of model performance degradation due to concept shift for users of ML models in domains like healthcare or time-series analysis, though it is incremental as it builds on existing methods like GAMs and knockoffs.
The paper tackles the problem of concept shift in machine learning by proposing SGShift, a model that detects concept shift in tabular data and attributes reduced performance to a sparse set of shifted features, achieving AUC >0.9 and recall >90%, often 2-3 times higher than baselines.
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.