This project is the culmination of a semsester's work in Machine Learning for Cybersecurity Analytics at Rochester Institute of Technology, Spring 2022.
The purpose of this project was to investigate what will create a better or more accurate machine learning model, supervised or unsupervised feature selection. The supervised feature selection method used was based on the XGBoost feature importances from SciKit Learn. This was used with the XGBoost Classifier. The unsupervised feature slection method used was PCA using the XGBoost classifier as well.
The results show that a supervised feature selection method can produce a much more accurate model while also fitting the XGBoost model much quicker using this IDS dataset.
The Jupyter notebook was created and ran in Google Colab.
The full problem description, proposed method, and results can be found in the PDF file included in this GitHub.
The data can be found at: https://research.unsw.edu.au/projects/unsw-nb15-dataset However, I have included the training and testing sets I used that I took out the categorical variables in this GitHub.