Research On Classification Method Based On Acupuncture Text Data

The research of artificial intelligence in the medical field has gradually become a hot spot of attention.The online medical treatment platform based on artificial intelligence-assisted is mostly an expert mode which consumes more energy,manpower and financial resources.In recent years,acupuncture treatment has been paid more and more attention because of its special treatment effect.Online medical treatment system which can give the disease for preliminary judgment automatically,recommend some relative acupuncture treatment plan and a self-directed way by patient symptoms is important,and the core technology of the system is construct a classification model of disease symptoms with higher accuracy.This paper uses machine learning and deep learning theory to solve expert mode problem of current online medical treatmen system by constructing a classification model for disease symptoms in acupuncture text data.Acupuncture text data is collected by hospital and internet which exist different characteristics from other common data sets,so it is necessary to analyze other short text classification methods when classifying the disease symptom data.The CHI is one of the most commonly used feature selection methods because of its low computational complexity when data sets less.However,the traditional CHI ignores the frequency of feature item in the essay,and there are also problems such as the negative correlation between feature items and categories.Based on this,combined with the inherent characteristics of acupuncture text data,this paper put forward a new hybrid feature selection method which is applies the TextRank algorithm extract the class keywords before using the CHI,and then extend reserved class keywords into the documentvector,this method can avoid the problems of the traditional CHI.At last combine with SVM classification algorithm establish a disease symptom classification model based on CHI's hybrid feature selection method.Deep learning algorithms are used in many fields widely.After extensive literature review,using CNN to make feature extraction and using SVM to classificate of disease symptom data has not yet appeared in the medical field.In this article,When there are many data sets,the short text is represented by word vector which be obtained by Skip-Gram training wiki Chinese corpus,and use CNN extract feature of short text.At last combine with SVM classification algorithm establish a disease symptom classification model based on CNN and SVM.At the same time,This model avoids the problem of high feature dimension and sparse data in traditional feature selection methods.In this paper,the proposed two disease symptom classification models were tested in a sample data sets.Through experimental results comparison,each of the established model is superior to other classification algorithms in the three evaluation indexes of accuracy,recall and F measurement.