首页|健康信息画像构建及虚假健康信息识别:融合社会感知数据与发布者先验知识

健康信息画像构建及虚假健康信息识别:融合社会感知数据与发布者先验知识

扫码查看
[目的/意义]融合包含丰富个体情感、行为和交互信息的社会感知数据和发布者先验知识有助于提高虚假健康信息识别精度.[研究设计/方法]基于社会感知数据,综合历史信息文本描述发布者对待检测信息的先验知识,并融合发布者先验知识,从发布者特征、内容特征和接收者行为特征3个维度提取健康信息特征;同时,建立健康信息画像,并基于Stacking集成学习模型构建虚假健康信息识别模型FHIR_SSD&PPK.[结论/发现]FHIR_SSD&PPK模型识别虚假健康信息的效果最好,准确率为92.35%;发布者特征的特征重要度占比总和最高,为51.59%,其中发布者先验知识特征的特征重要度为44.01%,并且与未考虑发布者先验知识的模型相比,F1值提升2.26%,说明本文提出的发布者先验知识是构建识别模型的关键特征.[创新/价值]FHIR_SSD&PPK模型融合社会感知数据和发布者先验知识,基于Stacking集成学习模型识别虚假健康信息,在细粒度和深度上对虚假健康信息识别研究进行了优化.
Construction of Health Information Portrait and the Identification of False Health Information by Integrating Social Sensing Data with Publisher's Prior Knowledge
[Purpose/Significance]The objective of this study is to explore how to integrate social sensing data,containing rich individual emotion,behavior,and interaction information,with publisher's prior knowledge to enhance the accuracy of false health information recognition.[Design/Methodology]Based on the social sensing data and historical information text,this paper describes the prior knowledge of publishers about detection information.By integrating the publisher's prior knowledge,the study extracts health information features from three dimensions:publisher features,content features,and receiver behavior features.Concurrently,health information portraits are established and the Stacking ensemble learning models is used to build a False Health Information Recognition Model(FHIR_SSD&PPK),a false health information recognition model that integrates social sensing data and publisher's prior knowledge.[Findings/Conclusion]FHIR_SSD&PPK model has the best effect in identifying false health information,with an accuracy of 92.35%.The total feature importance weight of the publisher features accounts for the highest proportion,at 51.59%,among which the feature importance weight of the publisher's prior knowledge features is 44.01%,and the F-Measure increases by 2.26%compared to the model without considering the publisher's prior knowledge,indicating that the publisher's prior knowledge proposed in this article is a key feature for building an identification model.[Originality/Value]The FHIR_SSD&PPK model integrates social sensing data and publisher's prior knowledge,identifies false health information based on the Stacking ensemble learning model,optimizing the research in fine granularity and depth.

Health information portraitFalse health informationSocial sensing dataPrior knowledgeStacking ensemble learning

赵又霖、庞航远、石燕青

展开 >

河海大学商学院,南京,211100

南京大学信息管理学院,南京,210023

南京农业大学信息管理学院,南京,210095

健康信息画像 虚假健康信息 社会感知数据 先验知识 Stacking集成学习

2024

图书情报知识
武汉大学

图书情报知识

CSTPCDCSSCICHSSCD北大核心
影响因子:1.649
ISSN:1003-2797
年,卷(期):2024.41(6)