How nagarjuna325/cs584-ml-project-spring23 works
Project Type and Purpose
• This project is a course project from CS 584 focused on applying machine learning techniques to detect anomalous network intrusions.
• It uses two Jupyter Notebook files—one for training and one for testing data—to perform feature selection, exploratory data analysis (EDA), and a suite of machine learning models (both supervised and unsupervised).
• The project includes dataset files (CSV and ZIP) containing the UNSW-NB15 dataset and feature description, which are essential parts of its data pipeline.
File Structure
• Top-level items include:
• Jupyter Notebook files: one for training and one for testing. These are the core codebases where the ML workflows are implemented.
• A dataset folder containing:
▪ NUSW-NB15_features.csv (feature descriptions)
▪ UNSW_NB15_testing-set.csv (testing dataset)
▪ UNSW_NB15_training-set.zip (training dataset in a compressed file)
• Additional documents (reports and proposals) which can be noted but are less critical for the technical architecture.
• There are no separate backend or frontend directories, so the entire focus is on data processing and ML modeling using Jupyter Notebook environments (both local e.g., Jupyter Notebook, and Google Colab).
README for Additional Insights
• The README highlights that there are two distinct workflows corresponding to the training and testing datasets.
• It clearly outlines the distribution of work among collaborators, hinting at the existence of both supervised (Logistic Regression, Decision Tree, SVM, XGBoost, Random Forest) and unsupervised models (MLP Classifier, CNN, RCNN).
• The README also notes that due to CPU/GPU constraints, the workflows are split; this implies an architectural decision to separate concerns between the training and inference/testing pipelines.
Final Checklist
• Every central component (datasets, notebook workflows, and ML model implementations) is visually distinct.
• Incorporate notes on split execution environments (local vs. cloud-based) if relevant to the project execution.
• Be very detailed: include sub-divisions (preprocessing, model training, evaluation) wherever appropriate to capture the overall ML pipeline.