Incremental Constrained Clustering by Minimal Weighted Modification
Résumé
Clustering is a well-known task in Machine Learning that aims at grouping instances according to their similarity. It is unsupervised and therefore difficult to tune without human input. In constrained clustering, subject matter experts (SME) can incorporate knowledge through constraints in the clustering algorithm. Involving an expert results in an incremental process where expert constraints are added on the fly, iteratively.
Although satisfying the constraints is crucial, successive partitions should look alike to avoid disturbing the expert. In this paper, we present an incremental constrained clustering framework integrating active query strategies and a constraint programming model to fit the expert's expectations while preserving cluster structure, so that the user can understand the process and apprehend its impact. Our model supports instance and group-level constraints, which can be relaxed. Experiments on reference datasets and a case study related to the analysis of satellite image time series show the relevance of our framework.
Origine | Fichiers éditeurs autorisés sur une archive ouverte |
---|---|
Licence |