Cybersecurity in social networks: An ensemble model for Twitter bot detection

Abstract

The increasing presence of bot accounts on social media platforms creates major challenges for ensuring truthful and reliable online communication. This study examines how well ensemble learning techniques can identify bot accounts on Twitter. Using a dataset from Kaggle, which provides detailed information about accounts and labels them as either bot or human, we applied and tested several machine learning methods, including logistic regression, decision trees, random forests, XGBoost, support vector machines, and multi-layer perceptrons. The ensemble model, which merges predictions from individual classifiers, achieved the best performance, with 90.22% accuracy and a precision rate of 92.39%, showing strong detection capability with few false positives. Our results emphasize the potential of ensemble learning to improve bot detection by combining the strengths of different classifiers. The study highlights the need for reliable and understandable detection systems to preserve the authenticity of social media, addressing the changing tactics used by bot developers. Future research should explore additional types of data and ways to make models easier to understand, aiming to further improve detection results.