基于文本挖掘技术的HIV/AIDS相关在线健康社区用户的健康需求分析

Analysis on health needs of users in HIV/AIDS related online health communities based on text mining technology

目的挖掘在线健康社区中HIV/AIDS相关板块的用户问答数据,分析其隐藏的健康需求。方法利用Python语言爬取截止于2023年8月10日的“患-患”型、“医-患”型在线健康社区信息,利用Jieba分词配合CNKI医学主题词表进行文本预处理后,使用t-SNE算法降维聚类并使用小批量K-means主题聚类获取详细信息,并通过TextCNN算法建立情感分析模型。采用轮廓系数对聚类效果进行评估。结果共纳入7个在线健康社区,采集258379条数据,词云图呈现分词所得40860个词语。高频词前10位分别为加油、吃药、希望、知道、感染、朋友、医院、对象、感觉和喜欢。t-SNE聚类轮廓系数共探测出3个主题类目,分别为健康信息需求、情感支持需求、社会陪伴需求,轮廓系数为0.72。其中健康信息需求涵盖疾病预防、诊断和治疗全过程。情感支持需求利用情感倾向分析图展示,该群体发布言论的正向情感支持倾向占比24.44%、负向情感宣泄倾向占比15.59%,同时中性词汇占比为60.97%。社会陪伴需求占比最高,可细分为线上交友、线下交友。结论在线健康社区是HIV感染者/AIDS患者、HIV感染高风险人群寻求社会陪伴、情感支持、健康信息的渠道之一,可进一步优化平台相关功能以提供开展以用户需求为导向的针对HIV/AIDS宣传教育、干预等防治工作的实施场地。

Objective To mine the question-and-answer data of users in HIV/AIDS(human immunodeficiency virus/acquired immunodeficiency syndrome)related sections of online health communities,and to to analyze their hidden health needs.Methods Python programming language was used to scrape information from the“patient-patient”and“doctor-patient”online health communities as of August 10,2023.After text preprocessed with Jieba segmentation and CNKI medical thesaurus,t-SNE algorithm for dimensionality reduction clustering and small-batch K-means topic clustering were used to obtain detailed information.A sentiment analysis model was established by TextCNN algorithm.The clustering effectiveness was assessed by silhouette scores.Results A total of 7 online health communities were included,and 258379 records were collected.The word cloud map represented 40860 words extracted after segmentation.The top ten high-frequency words were“cheer up”,“take medicine”,“hope”,“know”,“infection”,“friends”,“hospital”,“partner”,“feeling”and“like”.The t-SNE clustering with a silhouette score of 0.72 detected three thematic categories,namely,health information needs,emotional support needs and social companionship needs.Health information needs encompassed the whole process of disease prevention,diagnosis and treatment.Emotional support needs,as indicated by sentiment analysis chart,showed that positive sentiment support tendency accounted for 24.44%,negative sentiment catharsis tendency accounted for 15.59%and neutral words accounted for 60.97%among the comments from this group.Social companionship needs had the highest proportion and could be further categorized into online and offline friendship.Conclusions Online health community is one of the channels for HIV/AIDS patients and people at high risk of HIV infection to seek social companionship,emotional support and health information.Further optimization of platform-related functions can be carried out to provide a venue for the implementation of user-oriented HIV/AIDS publicity,education,intervention and other prevention and control work.