Uses Cersi (2017) to determine if a twitter account/tweet is a trollbot or not.
Based on https://arxiv.org/abs/1707.00086 by Emilio Ferrara (who has written a series of brilliant papers on Bots)
The data can easily be replaced to your own or you can add your own. The deep learning model has an accuracy of 98% in the training set and 97% in test set. Similary Adaboost classifier with SMOTE and ENN has a 100% acuracy in training set and 98.5% in the test set.
This repo currently only contains code for profile based detection. For tweet based, you can have a look at my commit histroy or attempt your own.
Currently this is a just notebooks that can be replicated. I can create a importable module if this gets enough interest.