Downloads: 1
Research Paper | Computer Science and Engineering | Volume 15 Issue 2, February 2026 | Pages: 1018 - 1026 | India
Social Media Bot Detection Using Account Metadata, Advanced Feature Selection, and Stacking
Abstract: The emergence of social media platforms, particularly Twitter, has transformed online communication while also introducing the issue of automated accounts, commonly referred to as social media bots. These bots have the potential to skew public conversations and spread false information, making their identification crucial for preserving online integrity. This research introduces a scalable and interpret-able machine learning framework designed to detect Twitter bots by analyzing only user account metadata, without relying on content-based features. The dataset includes over 37,000 accounts, with a slight imbalance favoring human accounts. The approach focuses on thorough data pre-processing and sophisticated feature engineering, such as calculating followers-to-friends ratios, activity levels, account age, and verification status, along with systematic feature selection using Random Forest, XGBoost, Recursive Feature Elimination (RFE), and Boruta. The most significant features consist-ently pertain to network attributes and account longevity. Various classifiers, including Decision Trees, Random Forests, XGBoost, and AdaBoost, are assessed, with ensemble models showing superior performance. Notably, Random Forest and XGBoost achieve ROC-AUC (Receiver Operating Characteristic Area under the Curve) scores exceeding 0.93 and F1-scores around 0.81. The stacking ensemble has boosted robustness, achieving an overall accuracy of 87% and an F1-score of 0.80 in identifying bots. This interpretable framework not only delivers high detection accuracy but also pro-vides valuable insights into behavior, facilitating effective adaptation to changing online threats.
Keywords: Social media, Twitter, Bot Detection, Recursive Feature Elimination, Stacking, Random Forest, XGBoost
How to Cite?: Shraddha R. Mehetre, "Social Media Bot Detection Using Account Metadata, Advanced Feature Selection, and Stacking", Volume 15 Issue 2, February 2026, International Journal of Science and Research (IJSR), Pages: 1018-1026, https://www.ijsr.net/getabstract.php?paperid=SR26216223812, DOI: https://dx.dx.doi.org/10.21275/SR26216223812