M4R

M4R Project

This repository only contains the code. So training data, collected data, trained model weights etc will not be included.

Contents

• Data Harvesting
• • tweet_harvester.py Harvests tweets based on search terms or tweet ids.
• • user_harvester.py Harvests tweets from specific users or user account data (specifically to fill out retweet and reply networks - i.e. to be able to perform account level detection).
• • full_text_tokeniser.py Tokenises the tweets.
• Bot Detection Methods
• • feature_selection.py Performs feature selection techniques, including RFC feature importances, recursive feature elimination, and ANOVA.
• • account_level_detection.py Trains account level detection model: compares models and resampling techniques. Also applies account level detection model to collected datasets.
• • tweet_level_detection.py Trains tweet level detection: compares models.
• Application to Election Data
• • data_exploration.py Explores summary statistics of the Georgia and US datasets, plots distributions of datasets, compares to training dataset.
• • sentiment_analysis.py Performs VADER sentiment analysis: distribution and plots over time.
• • reply_network.py Builds simple and more complicated reply networks for the Georgia dataset.
• • hashtag_cooccurrence_network.py Builds a hashtag co-occurrence network that can be plotted in Gephi.