Background:
With the rapid expansion of social networks and their transformation into primary
sources of news and information for millions of people, the issue of spreading
rumors and misinformation has become a serious challenge. Twitter, as one of the
main platforms for quickly sharing information, provides a suitable environment
for the spread of rumors. These rumors can have widespread negative impacts on
society, from creating public panic and undermining trust in social institutions to
influencing election outcomes and political decisions. Therefore, detecting and
combating rumors on this social network is of utmost importance.
Aim:
The aim of this research is to develop and evaluate methods for detecting rumors
on Twitter that can identify and prevent the spread of these rumors quickly and
accurately. These methods include sentiment analysis, identifying and limiting bots,
and using automated systems to evaluate the credibility of sources. The present
study seeks to provide a comprehensive framework that, by combining these
methods, can effectively reduce the spread of rumors and help maintain information
health in society.
Methodology:
The methodology of this research involves designing and implementing an
optimized model for detecting rumors in social networks, which utilizes a social
bot-aware graph model. First, the data used was collected from the Twitter 15 and
Twitter 16 datasets and employed as the input to the model. Then, four main
architectures, including Graph Attention Network (GAT), Graph Convolutional
Network (GCN), Textual Encoder, and Output Layer, were utilized, each with a
distinct role in data extraction and analysis. Additionally, a genetic algorithm was
employed to optimize the model's parameters to improve the accuracy and
efficiency of detecting rumors and bots in the network. Instead of multi-layer
perceptron models, the research utilized the Extreme Gradient Boosting Classifier
(XGBClassifier) for bot detection, and two new parameter