Twitter Bot Detection and Ranking using Supervised Machine Learning Models

The rise in use of social media has led to a sharp increase in the number of social bot accounts on platforms like Twitter. A majority of these accounts are used for malicious activities such as opinion mangling, hashtag spamming and directing users to spam URLs. These bots can cause societalunrest and manipulation if they are left unchecked and are not removed. This paper proposes a supervised machine learning approach with multiple models to not only detect these bots but also analyse the impact of these accounts to the community and rank them. Separate machine learning models are created to classify accounts based on user metadata, tweet content and posting history. The models are trained on the MIB dataset. A custom rule based algorithm is used to combine the predictions of these models to achieve an accuracy of 95.7%. This algorithm gives a better accuracy than any model gives in isolation.