Data clustering via cooperative games: A novel approach and comparative study

•We investigate a new clustering algorithm (HGC) based on hedonic games.•In HGC, data points are viewed as players that compose clusters (coalitions) according to a shared nearest-neighbor based matrix of preferences.•We compare HGC with popular clustering algorithms and with two methods also based on cooperative games (BiGC and DRAC).•Experiments were conducted on both low and high-dimensional data sets and the results were assessed by 10 external validation indices.•HGC is more effective and stable than BiGC and DRAC and usually converges fast to Nash stable data partitions.