Algorithm Design and Optimization for Knowledge Fusion: A Graph Matching-based Approach

Knowledge fusion plays a crucial role in the construction of knowledge graphs, aiming to integrate knowledge from different data sources and improve the accuracy and completeness of the knowledge graph. Knowledge graphs built from different data sources often suffer from issues such as inconsistency, redundancy, and missing information, which negatively impact the quality and application effectiveness of the knowledge graph. Therefore, how to effectively fuse knowledge from different data sources has become an important topic in knowledge graph research. This paper proposes a graph matching-based algorithm for knowledge fusion, aiming to merge knowledge graphs from different sources into a unified one. The algorithm achieves knowledge fusion by identifying and matching identical or similar nodes and edges in different knowledge graphs, involving several steps including node matching, edge matching, and conflict resolution. Node matching identifies identical or similar nodes by calculating the similarity of their attributes, while edge matching further matches the relationships between nodes based on node matching. Conflict resolution is handled using rules or statistical methods to address conflicting information from different data sources, ensuring consistency and accuracy in the fusion results. To improve the efficiency and effectiveness of the algorithm, this paper proposes a series of optimization methods: feature selection and weight allocation, iterative matching and fusion, as well as parallel computing. Feature selection and weight allocation optimize the similarity calculation process by selecting key features and assigning appropriate weights to improve the accuracy of node and edge matching. Iterative matching and fusion continuously optimize fusion results by gradually performing node and edge matching. Parallel computing utilizes parallel computing techniques to accelerate the fusion process of large-scale knowledge graphs, enhancing algorithm processing capabilities and efficiency. Experimental results demonstrate significant effectiveness of the proposed algorithm in knowledge graph fusion. By testing with multiple publicly available knowledge graph datasets, the algorithm shows significant improvements in node matching accuracy, edge matching accuracy, and overall fusion quality compared to traditional fusion methods. Specifically, the average node matching accuracy has increased by over 10%, edge matching accuracy has significantly improved, and the overall fusion quality is also superior to the comparative methods. The experiments validate that the graph-based knowledge fusion algorithm can effectively enhance the quality and accuracy of knowledge graphs, demonstrating wide application prospects and practical value.