A Shape-Based Method for Concept Drift Detection and Signal Denoising
Fabian Hinder, Johannes Brinkrolf, Valerie Vaquet, Barbara Hammer · 2021 IEEE Symposium Series on Computational Intelligence (SSCI) · 2021
The notion of concept drift refers to the phenomenon that the distribution, which is underlying the observed data, changes over time; as a consequence machine learning models may become inaccurate and need adjustment. Many unsupervised approaches in this context rely on measuring the discrepancy between the sample distributions of two time windows. However, due to the small number of samples that is used for the estimation, the results are usually rather noisy, which makes it hard to distinguish the actual, drift induced differences from the noisy ones induced by inner distribution variances. In this paper we analyze the structural properties of the drift induced signals. We derive a shape based filter mechanism, derive an efficient algorithmic solution, and demonstrate its usefulness in several experiments.