Credit Card Fraud Detection Notebook

Overview

This Jupyter notebook predicts credit card fraud using machine learning, exploring data, preprocessing imbalances, training models, and optimizing performance. It uses anonymized features (V1–V28), Time, Amount, and a binary Class label (0 for non-fraud, 1 for fraud).

Project Phases

1. Importing Required Libraries: Loads pandas, numpy, seaborn, matplotlib, sklearn, imbalanced-learn, xgboost, and metrics for analysis and modeling.
2. Load Data: Imports the creditcard.csv dataset containing transaction details.
3. Exploratory Data Analysis (EDA): Analyzes dataset shape, info, first rows, summary statistics, and class distribution to identify patterns.
4. Data Preprocessing: Engineers features (Hour, Amount_to_Mean, Hourly_Fraud_Rate, Log_Amount), scales features, and splits data into train, validation, and test sets with SMOTE.
5. Handle Imbalanced Data: Applies SMOTE to oversample the minority (fraud) class.
6. Model Training: Trains Logistic Regression, Random Forest, and XGBoost with cross-validation, ROC curves, and confusion matrices.
7. Hyperparameter Tuning: Optimizes Random Forest using RandomizedSearchCV for recall, with threshold tuning.
8. Model Evaluation: Evaluates the tuned Random Forest model on the test set with classification report, ROC-AUC, and confusion matrix.
9. Conclusion: Summarizes key features (V4, V11, Hourly_Fraud_Rate), performance, trade-offs, and next steps.

How to Clone

1. Clone the repository and navigate to the project folder:

   git clone https://github.com/rk0802p/cc-fraud-detection.git
   cd cc-fraud-detection