Self diagnosis model training method and device based on factor graph model

Abstract

The embodiment of the invention discloses a self diagnosis model training method and a self diagnosis model training device based on a factor graph model. The self diagnosis model training method based on the factor graph model comprises the following steps that professional knowledge feature data is extracted from collected professional knowledge data, and doctor-patient communication feature data is extracted from collected doctor-patient communication data, wherein corresponding relationships between diseases and symptoms are respectively stored in the professional knowledge feature data and the doctor-patient communication feature data; a sparse factor graph model including a hidden layer is built according to contents of the professional knowledge feature data and the doctor-patient communication feature data; the sparse factor graph model is subjected to transfer training by utilizing the professional knowledge feature data and the doctor-patient communication feature data until the parameters of the sparse factor graph model are totally converged. The self diagnosis model training method and the self diagnosis model training device based on the factor graph model simultaneously solve the problems of domain knowledge deviation and experiential data sparsity.

Description

technical field [0001] The embodiment of the present invention relates to computer data processing technology, in particular to a self-diagnosis model training method and device based on a factor graph model. Background technique [0002] Today, when the Internet is very developed, self-diagnosis systems that provide self-diagnosis services to users through the Internet are very popular. [0003] Existing self-diagnosis systems can be roughly divided into self-diagnosis systems based on professional knowledge and self-diagnosis systems based on empirical data. The self-diagnosis system based on professional knowledge provides self-diagnosis suggestions for users with reference to domain knowledge. The data it gives is supported by a mature theoretical knowledge system, so it is very authoritative. However, in the actual diagnosis process, there are often some situations that cannot be foreseen by domain knowledge, such as the patient's environment and the patient's own phy...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the embodiment of the present invention proposes a self-diagnosis model training method and device based on a factor graph model to simultaneously solve the problems of domain knowledge deviation and empirical data sparseness

Images