Research on human-like solution method for graph isomorphic mathematical reasoning based on knowledge graph

In the process of using knowledge graphs to assist deep learning in logical reasoning, there are problems with weak discriminative and generalisation abilities, as well as insufficient stability. A model for mathematical reasoning based on knowledge graph is proposed, which extracts classification features through graph isomorphism network and integrates the reverse to forward thinking approach to design a human like reasoning model for elementary mathematics. The innovation of the research lies in the use of a hierarchical structure in the design of the inference engine, which makes the logical layers relatively independent and highly modular, thereby improving computational efficiency. The results showed that the highest accuracy of the human like solution system constructed in the study was 94%, and the shortest time was 61.4 seconds. This indicates that the reasoning system can quickly and accurately solve elementary mathematics problems, providing a new method for education and teaching.