Twitter Bot Account Detection Using Supervised Machine Learning

Abstract

Twitter is a primary social media platform gaining popularity among social networking websites at an alarming rate. Twitter's popularity and relatively open nature make it an excellent target for automated programs known as bots, which are computer programs that run automatically. In addition to spamming, bots can be used for various purposes, such as inducing conversations to change the topic of discussion, modifying user popularity, contaminating materials to spread misinformation, and conducting propaganda. This study's goal was to provide a fresh perspective on estimating the possibility of an account being identified as a bot by applying Machine Learning algorithms to a variety of scenarios. Both Random Forest and XGBoost algorithms are used in this application. The inquiry began with exploratory data analysis to determine the current status of the dataset. Next comes the process of model engineering, which involves the steps of requirement gathering and specification, feature selection and optimization, hyperparameter tweaking, and algorithm benchmarking. The findings of this investigation suggest that the XGBoost algorithm outperforms Random Forest, with an accuracy of 0.8908 for XGBoost and 0.8762 for Random Forest.