Social Media-Based Research On Topic Identification And Correlation Topic Prediction In The Drug Field

Objective:In the face of massive and heterogeneous medical data resources,the emergence of knowledge discovery makes it possible for researchers to obtain effective knowledge and information from massive data.At the same time,with the rapid development of "social plus medical treatment",social media platform has become a new way for people to obtain medical information and spread health knowledge.Based on this,this paper studies the core topic identification and related topic prediction in the field of medicine and health from the perspective of social media.With the help of relevant methods,we can realize the early discovery and early research of the emerging topics in the field of medicine,and then broaden the perspective of scientific research and point out the direction of future research,which is of methodological and practical significance to grasp the research hotspots and future development trends in the field of medicine.Object:Microblog data related to type 2 diabetes drugs on Sina Weibo’s official media platform(hereinafter referred to as microblog).Methods:Comprehensive use of python programming language,co-occurrence analysis,LDA theme model,link prediction and other research methods,with the help of Origin,Ucinet,VOS viewer and other visualization software,Weibo data are statistically analyzed.According to the characteristics of Weibo data analysis,co-occurrence analysis is introduced on the basis of LDA theme model to improve the quality of theme generation.On this basis,the core subject words of the document are determined according to the law of twenty-eight.On this basis,the weighted / non-weighted theme co-occurrence network is constructed by calculating the co-occurrence intensity.Finally,a number of link prediction indicators are introduced,and each index is calculated by weighted and unweighted algorithms respectively,and the index with higher AUC value is selected as the optimal prediction index.On this basis,the related theme prediction analysis of the optimal index is carried out.Results of the study:Core topic identification analysis: 20 core topics are obtained through the analysis of the topic model.From the point of view of the subject content,it covers all aspects of the field of diabetes drugs,and analyzes the future development trend of this field from the perspectives of people’s drug use experience,national policy,pharmacodynamics,the development prospect of the drug industry,and so on.it mainly involves drug safety research,diabetes drug industry development,generic drugs,rational drug use,new indications,adverse reactions,drug treatment effects,combined therapy and other topics.About 74.1%of the core theme pairs in the field of diabetes drugs,their co-occurrence strength is lower than the average value in this field.Furthermore,it proves the necessity of related theme prediction research in the field of drug research from the perspective of social media.Related topic prediction analysis: from the perspective of social media,in the diabetic drug field theme co-occurrence network,the prediction effect of the non-weighted theme cooccurrence network is better than that of the weighted network,indicating that the intensity value between nodes does not positively promote the accuracy of the network prediction,and the optimal index is the Katz index of the unweighted algorithm,and finally predicted five related topics.Finally,this paper proves the reliability and accuracy of the experimental results from the point of view of scientific research literature and patents.Conclusion:Through the analysis of type 2 diabetes drug-related Weibo in Weibo data,20 core topics are obtained,such as drug safety research,development of diabetic drug industry,generic drugs,rational drug use,new indications,adverse reactions,drug treatment efficacy,combined therapy and so on.on this basis,the emerging topics in this field are predicted and analyzed from the point of view of national policy,drug industry,patient feelings,drug research and development.It can be seen that the topics that are most likely to create link opportunities in the future mainly include rational drug use and drug action mechanism research,pharmacodynamics and type 2 diabetes treatment options,diabetes drug industry development,new drug indications,diabetes and multi-disease combination therapy.