Ebrahim Sahafizadeh

This research introduces a model to enhance stance detection in social networks using machine learning models like SVM and BERT. Given the importance of understanding user stances in social and political analyses, this study aims to improve the accuracy of these models by integrating structural and textual information from communities. The objective is to boost the performance of stance detection models by utilizing textual structures of users within text-based networks. This research seeks to validate the hypothesis that incorporating community information can improve stance detection model accuracy. The statistical population of this study comprises a set of tweets with specific stance labels. For sampling, tweets related to particular topics and active users were selected. The modeling approach includes constructing a textual network from tweets, identifying communities within these networks, and applying baseline models like SVM and BERT. Structural community features, extracted through community detection algorithms such as the Louvain algorithm, are added to these models. The findings indicate that using community information leads to a significant improvement in stance detection accuracy. Combined models that incorporate both textual and community features exhibit more reliable and balanced performance compared to other models. The results demonstrate that considering the structure of text-based networks can effectively enhance stance detection model accuracy, underscoring the importance of integrating text network community detection information alongside textual features for machine learning models in social analysis.