Skip to content

ML From Scratch

Back to Projects
ML From Scratch preview
Python Machine Learning C4.5 SVM
View Repository

Overview#

This project implements machine learning models without relying on high-level training APIs. The focus is understanding algorithm mechanics, not only model usage, and evaluating them on the UNSW-NB15 cybersecurity dataset.

Highlights#

  • Decision Tree C4.5 with split logic and pruning paths
  • Logistic Regression training via gradient-based optimization
  • SVM implementation and comparison workflow
  • Evaluation pipeline and reproducible experiment structure

Challenges & Learnings#

  • Getting numerical behavior stable in iterative training
  • Handling data preparation for realistic datasets
  • Comparing custom models with established baselines

Contributors#