Social Bot Detection

Social bots on Twitter are accounts controlled by software that can manipulate opinions in important events such as elections. In order to detect these bots, we want to achieve both scalability and generalizability, to enable faster real-time detection algorithms and to make the classification more robust to new bots that are different from previous ones. We replicate the paper Scalable and Generalizable Social Bot Detection through Data Selection (Yang, Varol, Hui, Menczer 2020) and test the robustness of its results. The paper proposed an approach for classification through training random forest models on just user profile information and only on a subset of the available training data. Our results align with Yang, et al. (2020) in that training on subsets of datasets with user metadata improves classifier performance. We differ from their models as we achieved better model performance than their top models using a fewer number of datasets for training. These differences might be due to datasets differences, feature calculation discrepancies, API call rate limit, and our relatively newer sampling time for account features of certain datasets. Relationships across all training and testing datasets are investigated to explain top models’ selections of training subsets, and feature importance in the training datasets are analyzed