Experimental comparison of heuristic cluster-editing algorithms for entity deduplication.

Abstract

The application of heuristic weighted cluster-editing algorithms within the scope of entity deduplication is a relatively unexplored area. This research has aimed at comparing the efficacy of different heuristics on both real-world and artificiallygenerated entity-deduplication data-sets. The research has shown that the Force, Spectral, Vote/BOEM, and Split-Merge heuristics perform relatively well for precision in comparison to the benchmark heuristics Pivot and Closure on a variety of data-sets