Feature Drift Detection in Evolving Data Streams

Abstract

Most current stream mining techniques can adapt to data distribution changes, known as concept drift. Common concept drift detectors focus on detecting and signaling drift when a model’s prediction accuracy deteriorates. To allow us to evaluate a model’s accuracy we need data with ground truth. We focus on feature drift that shifts the model’s boundaries, and present a framework to detect feature drift without labels. The framework detects abrupt and gradual feature drift by two distance functions, Wasserstein and Energy, and discuss feature changes in the data stream. A less explored area is describing the changes in the data stream. Crucially, the ability to describe changes in the data stream would enable a better understanding of the changing dynamics in the relationships that take place over time. In particular, we seek to answer the following question: Whether the distribution changes of important features will also cause concept drift. Experimental results show that the proposed framework detects and describes the feature drift.