When creating a model using machine learning the availability of labelled instances is vital. Once the model has been built their importance decreases as long as the concept remains static. However if the concept changes labelled instances are once again imperative. The focus of this research is to minimise the need for labelled instances when tackling concept drift.
The project is funded by ABBEST
Drift Detection using Uncertainty Distribution Divergence: please contact author: [escapeemail email="email@example.com"] for the two text datasets used.