A Headache Diagnosis Model Based On Machine Learning

Headache is a common neurological disease.How to make full use of the existing text data of Chinese headache cases(doctors’ experience)to assist headache diagnosis is an impor-tant topic in the field of intelligent medicine.Based on machine learning techniques such as ensemble learning and deep learning,this thesis studies headache features extraction methods and headache diagnosis methods for Chinese headache case texts,providing a new way for computer-aided diagnosis in headache field.The main researches of this thesis are as follows:1.In order to support headache features extraction,a normalized representation framework of headache attributes is designed? According to the normalized representation framework of headache attributes,the attributes of Chinese headache case text data are labeled,and a labeled corpus data set for headache features extraction is constructed.2.A headache features extraction model for Chinese headache case texts is designed.The open source Chinese vector thesaurus is used to express the Chinese character sequence vectorically? The extraction models include the headache features extraction model based on Bi-LSTM + CRF,the headache features extraction model based on BERT + Bi-LSTM + CRF,and the headache features extraction model based on IDCNN + CRF.The effectiveness of the extraction model is verified by experiments.Experimental results show that the headache fea-tures extraction model based on IDCNN + CRF has the best extraction effect and can provide more reliable headache features for headache diagnosis.3.A headache diagnosis model based on ensemble learning is designed,including the headache diagnosis model based on Bagging(KNN-based learner),the headache diagnosis model based on random forest,the headache diagnosis model based on Adaboost,and the headache diagnosis model based on Gradient Boosting.The effectiveness of the diagnosis model is verified by experiments.The effectiveness of the diagnosis model is verified by ex-periments.Experimental results show that the headache diagnosis model based on Gradient Boosting has the best headache diagnosis effect.4.With the headache features extraction model based on IDCNN + CRF and the headache diagnosis model based on Gradient Boosting as the core,a headache diagnosis prototype system based on ensemble learning is designed and implemented.The main contribution of this thesis is to propose a headache diagnosis method based on ensemble learning for Chinese headache case texts.This method has reference value for com-puter aided diagnosis in headache field.