Twitter Spam Accounts Detection Using Machine Learning Models

The amount of spam accounts on Twitter has recently surged, which has attracted researchers' interest in seeking strategies to mitigate this problem. This paper reviews recent studies in the literature that tackled the Twitter spam accounts problem based on machine learning (ML). It then introduces an empirical study to test several ML models on a publicly access dataset. The model types were individual, ensemble, and majority voting models. It found that the ensemble ML models, and majority voting ML models can improve the prediction accuracy of Twitter spam accounts detection compared to the individual ML models. We concluded that the Random Forest model is the best for Twitter spam accounts detection using an imbalanced dataset.