This project uses a Naive Bayes model to predict what Wikipedia articles are likely to be the subject of an edit war.
This project uses the Python Natural Language Toolkit, Scikit-learn, and the Requests library. I used a Naive Bayes model to predict edit wars based on word frequencies, word counts, and density of references.
This was my final project for my college machine learning class.
final_submission.pdf is a document explaining this project in-depth.
If you're more interested in looking directly at the code, you can look in the
src/ directory, which contains the following files:
scraper.py: This downloads articles from Wikipedia and extracts the contents out of the HTML.analyze_text.py: This parses the raw content of a wikipedia page. It filters out stopwords, lemmatizes the remaining words, and then converts the result into a "bag-of-words".calculate_controversy.py: This calculates articles' controversy scores using the method described in this paper.naive_bayes.py: This runs the actual machine learning algorithm which predicts how likely it is that an article will have an edit war.final.py: This file contains miscellaneous utility functions used throughout the other Python files.