UNSW-NB15 ML Preprocessing Pipeline

This repository provides a clean, reproducible preprocessing workflow for the UNSW-NB15 dataset. It merges training/testing CSVs, cleans features, encodes labels (binary + multiclass), and generates stratified Train/Validation/Test splits ready for machine learning models.

Features

• Merge + clean UNSW-NB15 training/testing sets
• Remove metadata noise (id column)
• Replace NaN/Inf and enforce numeric features
• Binary & multiclass label encoding
• Stratified 70/15/15 Train–Val–Test split
• Saves all outputs into /processed-data/

Project Structure

raw-data/
├── UNSW_NB15_training-set.csv
└── UNSW_NB15_testing-set.csv

processed-data/
├── X_train.csv
├── X_val.csv
├── X_test.csv
├── y_train.csv
├── y_val.csv
└── y_test.csv

UNSW-NB15 Preprocessing & Train_Val_Test Split.pdf

UNSW-NB15_Notebook.py
notebook_output.txt

How to Run

python UNSW-NB15_Notebook.py

Processed splits will be saved in processed-data/.

Dataset

UNSW-NB15 original dataset is available from UNSW Canberra Cyber - https://research.unsw.edu.au/projects/unsw-nb15-dataset.

---