Knowledge map reasoning algorithm based on Weight

Auxiliary triage is an important direction for the application of medical knowledge graphs. Entity reasoning is the technical difficulty to realize auxiliary triage. The existing methods mainly rely on entity reasoning algorithms based on subgraph extraction, which have the disadvantages of large calculation amounts and low accuracy. This paper proposes a weight-based entity reasoning algorithm. The method uses Naive Bayes to calculate the weight of the relationship between entities, and quantifies the local structural characteristics of the entity by calculating the purity of the entity and the correlation between the entities, and based on this to improve the subgraph extraction process to reduce the complexity of the algorithm. The experiment is based on the open-source medical knowledge map of the Chinese Academy of Sciences and the patient data set provided by the partner hospital. The average time consumption is 0.252s, the additional storage consumption is 950KB, and hits@5 is 91.34%. At the cost of occupying a small part of the storage space, inference speed and accuracy have been improved.