Automatic Entity Matcher Combination for Heterogeneous Entity Alignment via Multi-chromosome Genetic Algorithm

Ontology Matching (OM) plays a critical role in enabling seamless data interoperability and automated reasoning by discovering semantic correspondences between heterogeneous ontologies. Because no single Entity Matcher (EM) can fully capture lexical, structural, and semantic variations among diverse ontologies, aggregating multiple EMs is essential to achieve accurate matching results. To enhance the effectiveness and efficiency of OM, we propose a novel Multi-chromosome Genetic Algorithm (MGA), which includes three new components. First, a new multi-chromosome encoding mechanism is designed to simultaneously optimize aggregation weights and thresholds, thereby enhancing matching accuracy. Second, to improve search efficiency, a novel breeding crossover operator is developed to capture complex interrelations among multiple chromosomes within an individual. Finally, a multi-chromosome local search strategy is presented to refine elite solutions to further boost optimization performance. Extensive experiments conducted on the Ontology Alignment Evaluation Initiative (OAEI) Benchmark and Conference datasets demonstrate that MGA consistently outperforms state-of-the-art methods. Specifically, MGA achieves substantial improvements in f-measure scores, achieving an average f-measure of 0.87 on the Benchmark dataset and 0.70 on the Conference dataset, and outperforming the best matching methods by 2.5% and 3.8%, respectively, which confirms its robustness and effectiveness in matching heterogeneous ontologies.