A Change Dynamic Model for the Online Detection of Gradual Change
This work addresses the challenge of gradual change detection in fields like medical monitoring, offering improved accuracy and speed, though it is incremental as it builds on existing change-point models.
The paper tackles the problem of detecting gradual changes in stochastic processes, where traditional change-point models assume instantaneous transitions, by introducing a novel change-dynamic model in a Bayesian framework. The result shows that this model enables faster and more accurate identification of gradual change in synthetic data and EEG readings during epileptic seizures compared to traditional methods.
Changes in the statistical properties of a stochastic process are typically assumed to occur via change-points, which demark instantaneous moments of complete and total change in process behavior. In cases where these transitions occur gradually, this assumption can result in a reduced ability to properly identify and respond to process change. With this observation in mind, we introduce a novel change-dynamic model for the online detection of gradual change in a Bayesian framework, in which change-points are used within a hierarchical model to indicate moments of gradual change onset or termination. We apply this model to synthetic data and EEG readings drawn during epileptic seizure, where we find our change-dynamic model can enable faster and more accurate identification of gradual change than traditional change-point models allow.