The invention relates to an epidemiological investigation method and
system based on a large
language model, and belongs to the technical field of
public health information.The epidemiological investigation method comprises the steps that a structured
knowledge base is constructed, and preprocessing, structured disassembly and
feature extraction are conducted on multiple types of
field data to form standardized data; fusing the processed data with a
knowledge graph of a
knowledge base, mining potential associated clues, inferring
pathogenesis and propagation paths through a model, constructing the
knowledge graph based on the
knowledge base, developing an associated clue mining and
pathogenesis inferring
algorithm, training and optimizing the model through case data, and forming the
knowledge graph covering event elements and causal relationships. According to the method, implicit experience dominance is realized, grassroots flow dispatchers can also achieve the clue capturing ability of qualified experts, and the working gap of flow dispatchers with different experience levels is reduced. The response speed is increased from the hour level to the second level for retrieval matching, the emergency disposal requirements of
public health emergencies are met, and time is won for epidemic prevention and control.