• Research

Scalable Active Temporal Constrained Clustering - Proceedings of the 21st International Conference on Extending Database Technology (EDBT), March 26-29,2018

Son T. Mai, Sihem Amer-Yahia, Ahlame Douzal Chouakria
doi: 10.5441/002/edbt.2018.44


We  introduce  a  novel  interactive  framework  to  handle  bothinstance-level and temporal smoothness constraints for cluster-ing large temporal data. It consists of a constrained clusteringalgorithm which optimizes the clustering quality, constraint vio-lation and the historical cost between consecutive data snapshots.At the center of our framework is a simple yet effective activelearning technique for iteratively selecting the most informativepairs of objects to query users about, and updating the clusteringwith new constraints. Those constraints are then propagatedinside each snapshot and between snapshots via constraint in-heritance and propagation to further enhance the results. Experi-ments show better or comparable clustering results than existingtechniques as well as high scalability for large datasets.

Published on August 22, 2018