Research On Knowledge Discovery Based On Text Mining In Online Medical Community

The booming development of internet medical service has effectively alleviated the long-standing problems of China’s lack of high-quality medical service supply,limited and uneven distribution of physician resources.As a product of internet medical service,online medical community extends medical services to the internet environment,which brings about a new development of medical service mode.It plays an important role in reducing medical costs,improving the allocation efficiency of medical resources and enhancing the fairness of medical resources.Currently,online medical community has gathered hundreds of millions of users,accumulating massive amounts of text data about patient consultations & diagnoses,and publicly available text data from doctors.Those text data contain a wealth of implicit,previously unknown,and potentially valuable rules and information.In-depth mining and analysis of information and knowledge hidden in online medical text data,identification of hot topics for patient consultation,exploration of regional differences and evolutionary trends in medical resources,analysis of flow direction of patients seeking medical service and discovering the potential comorbidity between diseases are of great practical significance for improving the quality of medical services,optimizing the allocation of medical resources,improving the medical supply mode,and strengthening disease prevention and control measures.Throughout the existing researches,scholars have extensively explored knowledge discovery research on online medical communities.However,due to the difficulty of effectively extracting knowledge from online medical community texts,there are still important but not yet fully researched issues in the research of knowledge discovery in online medical communities.Therefore,on the basis of relevant basic theories and online medical research literature review,this paper carries out research on online medical community knowledge discovery from the following four aspects:(1)At present,most online medical communities do not have a clear topic organization structure,and information on various topics is scattered in the online medical communities,which makes it difficult for patients to quickly retrieve the required information,and doctors cannot accurately grasp the real needs of patients.Clarifying patients’ needs plays a crucial role in improving medical services,carrying out precision treatment,and easing doctor-patient relationship.Therefore,the first knowledge discovery research of this paper is to identify the hot topics of patient consultation in online medical communities.(2)Doctors are the core resources of the medical system,and the uneven distribution of doctor resources will inevitably increase the phenomenon of patients seeking medical treatment in different places,increase the cost of patients seeking medical treatment,thus hindering the realization of health equity and deviating from the goal of balanced regional development.Accordingly,the second knowledge discovery research of this paper explores the provincial regional differences and stage evolution trend of medical resources from the perspective of doctors’ professional skills.(3)Depression is becoming the second largest killer of human beings,second only to cancer.Depression can significantly affect physical and mental health,cause inner emotional instability,and even cause suicidal thoughts in severe cases.Although many methods can successfully cure depression,less than 25% of depressed patients receive effective treatment,mainly due to the lack of mental health resources and the stigma brought by mental illness.Consequently,the third knowledge discovery research of this paper aims to identify the condition of depressed patients in online medical communities,explore the flow direction of seeking medical service and other characteristics of patients with depression.(4)The existence of comorbidity makes the combination of multiple chronic diseases cause harm to the human body,which is more difficult to treat and judge than a single disease.Chronic diseases in the mode of comorbidity have become the biggest threat to patients.As diabetes is a serious chronic disease burden in the world,it is self-evident that it is important to explore the potential comorbidity of diabetes.Therefore,the fourth knowledge discovery research of this paper is the discovery of the potential comorbidity of diabetes.To quickly and accurately mine and discover the above valuable specific knowledge from such a large amount of text data in online medical community,it is necessary to rely on intelligent technological means to address the challenges brought by the information explosion.The current research on knowledge discovery of text data in online medical communities mostly adopts text mining methods in the general field.On the basis of various semantic information of general text,the text of online medical community also contains rich knowledge in the medical field.The applicability of text mining methods in the general field to texts in online medical communities still needs to be verified,and the accuracy needs to be further improved.Currently,there is no systematic academic research on text mining methods for knowledge discovery in online medical communities,and its theoretical basis is relatively weak.Thus,in order to make more comprehensive and in-depth use of text data in online medical communities and promote the high-quality development of online medical big data applications,this paper focuses on the key methods of text mining to conduct research on knowledge discovery in online medical communities,the main work and research contents are as follows:(1)For the research on discovering hot topics in online medical community,this paper innovatively proposes a text similarity measurement method according to the characteristics of text data in online medical community,and tentatively applies the density peak clustering algorithm to the research on the discovery of hot topics in online medical community.The similarity between texts is converted into the sample point distance and local density involved in the density peak clustering algorithm to avoid errors caused by secondary calculation.This research implements the effective clustering of patient consultation text data in online medical community based on the density peak clustering algorithm,and discovers the hot topics commonly concerned by patients in the department of endocrinology in online medical community.Finally,different characteristics of different hot topics are also found by the visual method.(2)For the research on regional differences and evolution trends of medical resources in online medical community,this paper starts from the perspective of doctors’ professional skills,aims at the specialty text data of doctors who have opened personal homepages in online medical community,constructs a BERT-Bi LSTM-CRF named entity recognition model based on deep learning,extracts two types of entities in the text,i.e.diseases and treatment methods,and proposes a measure method of doctors’ expertise level in different regions and at different times.Taking cardiovascular medicine as an example,this paper conducts spatial visualization analysis on the measurement level of doctors’ expertise in different provincial regions,and then analyzes the evolution of the measurement level of doctors’ expertise in different years according to the time when doctors settled the platform.Based on the text mining method,this research provides a feasible strategy for analyzing the differences in specialty among doctors,and provides a reference for exploring the spatial heterogeneity of medical resources distribution in China and the evolution trend of medical resources in online medical communities.(3)For the research on knowledge discovery related to depressed patients in online medical community,this paper aims at the symptom description text of depressed patients in online medical community,defines the text classification system according to the international classification standard of depression,and divides the symptom description text into four types: no depression tendency,mild depression,moderate depression,and severe depression.Three text classification models based on BERT pre-trained language model are constructed,and the model with the best performance is used to predict the condition of patients in online medical community based on the symptom description text.The flow direction of seeking medical service and other characteristics of patients with different degrees of depression in online medical community are analyzed by visual method.This research constructs the high-performance text classifier based on patient symptom description,aiming to find more useful information and knowledge about patients with different degrees of illness,and create more possibilities for platforms to provide medical services more accurately and doctors to improve differentiated online diagnosis and treatment.(4)For the research on discovering potential comorbidity in online medical community,this paper selects the association rule mining Apriori algorithm as the main research method.In order to solve the problem that Apriori algorithm needs to scan the whole database when calculating the support count of each candidate item set,which leads to a lot of time and space consumption,this paper proposes an improved Apriori method based on sparrow search algorithm.In this method,the Logistic chaotic map is used to initialize the sparrow population,so as to increase the diversity of sparrows and enhance the optimization ability of the algorithm.At the same time,an optimization strategy to reduce the search space is used to further improve the efficiency of the method.By comparing with the original Apriori algorithm and Apriori algorithm optimized by other swarm intelligence algorithms,the superiority of the proposed method in generating association rules in terms of quality and efficiency is verified.Finally,the proposed method is used to discover the potential comorbidity of diabetes in online medical community.