Research Article
On the Equivalence Between Community Discovery and Clustering
@INPROCEEDINGS{10.1007/978-3-319-76111-4_34, author={Riccardo Guidotti and Michele Coscia}, title={On the Equivalence Between Community Discovery and Clustering}, proceedings={Smart Objects and Technologies for Social Good. Third International Conference, GOODTECHS 2017, Pisa, Italy, November 29-30, 2017, Proceedings}, proceedings_a={GOODTECHS}, year={2018}, month={3}, keywords={Community discovery Clustering Problems equivalence}, doi={10.1007/978-3-319-76111-4_34} }- Riccardo Guidotti
Michele Coscia
Year: 2018
On the Equivalence Between Community Discovery and Clustering
GOODTECHS
Springer
DOI: 10.1007/978-3-319-76111-4_34
Abstract
Clustering is the subset of data mining techniques used to agnostically classify entities by looking at their attributes. Clustering algorithms specialized to deal with complex networks are called . Notwithstanding their common objectives, there are crucial assumptions in community discovery – edge sparsity and only one node type, among others – which makes its mapping to clustering non trivial. In this paper, we propose a community discovery to clustering mapping, by focusing on transactional data clustering. We represent a network as a transactional dataset, and we find communities by grouping nodes with common items (neighbors) in their baskets (neighbor lists). By comparing our results with ground truth communities and state of the art community discovery methods, we show that transactional clustering algorithms are a feasible alternative to community discovery, and that a complete mapping of the two problems is possible.