A Multi-Perspective Approach Utilizing Supervised Machine Learning Algorithms for Effective Social Media Bot Detection

Abstract

This research presents a comprehensive exploration of social media bot detection, focusing on Instagram, Twitter, and TikTok platforms. Social bots are autonomous machines that create for social media platforms. In this work present supervised machine learning algorithms, including models Random Forest, XGBoost, KNN, Decision Tree, and Neural Network, I am conducted an extensive analysis of bot behavior. Collect all public available datasets of labeled human and bot accounts. And through this study we have analyzed the behavioral activities of different bots accounts. The study aimed to understand the impact of multiple bot accounts, enhance accuracy through multi-account analysis, and deploy supervised machine learning technologies for improved platform security. In this research evaluated algorithms performance using key metrics such as accuracy, precision, recall, and F1-score, considering platform-specific challenges. Random Forest and XGBoost three-type accounts are consistently robust performers. Random Forest Instagram accuracy is 95%, Twitter accuracy is 91%, and TikTok accuracy is 97%, with the highest prediction accuracy. XGBoost Instagram accuracy is 95%, Twitter accuracy is 90%, and TikTok accuracy is 97%. All models achieve high accuracy, emphasizing the effectiveness of bot detection algorithms on the TikTok social media platform. The Neural Network maintains stable results with a balanced precision-recall trade-off, showing flexibility. Overall, the study provides valuable knowledge about social media bot detection, highlighting Random Forest and XGBoost as reliable choices. It gives strong user privacy and data security and opens the way for future research, inspiring exploration into ensemble approaches and advanced techniques.