Banknote Authentication Analysis

Binary classification benchmark authenticating genuine vs. forged banknotes using continuous wavelet transform features.

---

Overview

• Task: Supervised binary classification on the UCI Banknote Authentication dataset.
• Dataset: 1,372 records, 4 continuous numerical features, zero missing cells.
• Evaluation: Stratified cross-validation and holdout evaluation against baseline models.
• Target: Authentic (0) vs. Forged (1).

---

Data Summary

• Source: UCI Machine Learning Repository — Banknote Authentication
• Features:
  • variance: Variance of Wavelet Transformed image
  • skewness: Skewness of Wavelet Transformed image
  • curtosis: Curtosis of Wavelet Transformed image
  • entropy: Entropy of image
• Quality Check: 1,372 total records, 0 missing values, zero duplicates removed.

---

Visualizations

Data Quality & Target	Feature Distributions

Feature Correlations	Model Comparison

---

Repository Structure

├── figures/                   # Diagnostic, distribution, and evaluation plots
├── analysis.ipynb             # Interactive walk-through and evaluation notebook
├── audit.json                 # Run hashes and reproducible environment metadata
├── data_dictionary.csv        # Feature types, boundaries, and descriptions
├── descriptive_statistics.csv # Summary distribution stats per feature
├── feature_importance.csv     # Permutation importance rankings
├── metrics.json               # Full CV and holdout validation scores
└── README.md