Application Research On Recommendation Algorithm Based On Complex Network For Traditional Chinese Medicine Assisted Inquiry

As a traditional method of syndrome differentiation,Traditional Chinese Medicine(TCM)interrogation can estimate the most possible syndrome type according to the existing symptom information,and then find more symptom information to support the corresponding syndrome type,which leads to the question of inquiry.According to the new symptom information,we can further identify the symptoms and find more comprehensive symptom information until accurate diagnosis.In fact,the inquiry process is actually a dialectical process,whose result is limited by the doctor’s experience and medical level.The dialectical experience is often the most important factor for judging the ability of Chinese medicine diagnosis.Especially for the doctors lacking clinical experience,it is difficult to accurately and efficiently determine the symptoms of each step in the process of inquiry,resulting in collecting too much unrelated symptoms information.Symptoms information not only affects the efficiency of inquiry,it also increases the difficulty of differentiation and reduces the dialectical accuracy.Complex networks can be employed to solve the above problems because it can intuitively describe the complex relationship between symptoms.In addition,their community structure characteristics can well describe the relationship between symptom groups and syndrome types.In view of the characteristics of complex network,the recommendation method based on the complex network can obtain the symptoms of the inquiry through the analysis of the Chinese medical inquiry data to improve the accuracy of the inquiry.In order to improve the accuracy of the recommendation of the content,a recommendation algorithm based on the symptom related network is proposed to improve the accuracy of the recommendation.Our main work is as follows:(1)In order to discover the relationship between symptoms and syndromes in TCM inquiring data,we proposed the OCDSAN algorithm,which is based on overlapping community discovery.The algorithm uses the correlation between symptoms to construct an undirected initial symptom network.Considering the overlap between the symptoms in different syndromes,the node trust is defined for selecting the seed nodes to improve the degree of the symptom area.The neighbor nodes are extended from the seed nodes.In this way,all the overlapping communitiesare present and the symptom association network is constructed.The experimental results on real data sets show that the OCDSAN algorithm has good stability and high overlapping module degree on different scale networks,and the EQ value is increased by from 9% to 14% compared with the other two algorithms,and the association between symptoms and syndrome types is found effectively.(2)With the purpose of providing interrogation prompts for doctors based on symptom information,a query prompting recommendation algorithm(RASAN)based on symptom correlation network is proposed.The algorithm synthetically considers the impact of symptoms on different syndromes and defines the degree of community belonging,which can estimate the largest possible syndrome set of the patient.What’s more,the importance of node can measure the importance of each symptom in the community.Based on random walk in different types of communities,the RASAN algorithm forms a list of inquisition hints.Experiments on real data sets show that the accuracy and recall rate of the RASAN algorithm is increased by from 5% to 10%compared with other algorithms.This algorithm can provide effective hints for doctors.(3)This paper designs and implements an auxiliary inquiry recommendation prototype system in order to verify the feasibility of the proposed algorithm in the actual inquiry scene.This system can provide accurate and targeted inquisition tips for doctors,and offer the dialectical results according to the collected patient information.