feat(python-ml): EDA and preprocessing for credit card fraud detection dataset

[khushi09-oss]

Overview

Adds exploratory data analysis (EDA) and preprocessing for the Kaggle Credit Card
Fraud Detection dataset, laying the foundation for the fraud detection ML pipeline.

What's in this PR

• eda.ipynb - Jupyter notebook covering full dataset exploration and preprocessing:
  • Dataset overview: 284,807 transactions, 31 features (V1–V28, Time, Amount, Class)
  • Class imbalance analysis: 99.83% legitimate vs 0.17% fraud (577:1 ratio)
  • Amount & Time statistics broken down by class
  • Feature correlation analysis with fraud label (top signals: V17, V14, V12)
  • Normalization of Amount and Time using StandardScaler
  • Final preprocessed shape ready for model training

Key Findings

• Dataset is clean -> zero missing values
• V1–V28 are already PCA-scaled; only Amount and Time needed normalization
• Fraud transactions have lower median amount ($9.25) vs legitimate ($22.00)
• Accuracy is a misleading metric here -> Precision/Recall/F1 must be used