Credibility analysis of web based health information

Abstract

A growing number of people are using the Internet to address health issues, which has changed the traditional relationship between doctors and patients. Significant misunderstandings of health information have spread across the Internet and become an emerging public health concern. Many research and analysis works have been carried out in the field of false news identification, but few of them are intended to deal with challenges in web-based health information (WHI). Therefore, detecting the credibility of WHI is a critical problem that requires more attention. Amid this landscape, it is essential to recognize the dynamic changes in healthcare and the pressing need for alternative approaches in designing evaluation models. Adapting these models to assimilate vast multilingual databases becomes crucial in the face of evolving complexities. Moreover, the surge in WHI seekers necessitates the development of advanced interfaces supported by efficient algorithms. In this thesis, after conducting the in-depth literature review and assessing the intervention used in the field of WHI credibility, we employed exploratory factor analysis to identify specific criteria for assessing health information credibility from online sources, introducing four significant factors that integrate with social media engagement features to create a unique health misinformation dataset. Our work goes beyond existing solutions for countering online health misinformation, introducing a novel method using Autoregressive Honey Badger Optimization-based Deep Maxout Network (AHBO-based DMN) for credibility assessment. AHBO is devised by inheriting CAViaR features with HBA. The extraction of optimal features makes the classifier return the best solution. Addressing the challenge of limited and outdated datasets, we developed algorithms Tweet Annotation and Curation Algorithm (TACA) and Verifiable Information Gathering Algorithm (VIGA) to construct a multilingual health misinformation dataset. The factors derived from the survey results play a crucial role in devising the dataset. The performance showcases the superiority of our proposed model (AHBO-based DMN) over other traditional machine learning, deep learning, and nature-inspired algorithms in experimental results. We created a user-friendly medical chatbot that simplifies interactions with complex machine learning models, enhancing accessibility for users of all technical levels. The dataset constructed in this study can be used by other researchers as a benchmark dataset for the WHI domain, allowing them to compare and assess the effectiveness of their own models. Future studies should also examine the usage of self-attention models with different word embedding methodologies in deep learning architectures for user interest discovery and suggestions for refining the outcomes. In the realm of web-based health information (WHI) credibility assessment, our study introduces novel methodologies that address significant shortcomings in existing approaches. While previous research has focused on false news identification, our work pioneers a specialized approach tailored to the complexities of WHI. We employ exploratory factor analysis to define specific evaluation criteria, integrating social media engagement features to create a unique dataset for health misinformation analysis. Furthermore, we introduce the Autoregressive Honey Badger Optimization-based Deep Maxout Network (AHBO-based DMN) for credibility assessment, surpassing traditional machine learning and deep learning algorithms. Overcoming the limitations of existing datasets, our algorithms construct a comprehensive multilingual dataset, establishing a benchmark for future research in the WHI domain. Additionally, our user-friendly medical chatbot enhances accessibility for users of varying technical expertise, simplifying interactions with advanced machine learning models. By addressing these shortcomings and introducing innovative methodologies, our study contributes significantly to the advancement of WHI research, offering new insights and solutions for combating health misinformation in online environments.