💳 Credit Card Fraud Detection Using Machine Learning

A supervised machine learning project to detect fraudulent credit card transactions using Random Forest Classification on a highly imbalanced dataset.

---

Table of Contents

• Project Overview
• Dataset
• Project Structure
• Tech Stack
• Workflow
• Key Results
• Visualizations
• How to Run
• Possible Improvements
• Author

---

📖 Project Overview

Credit card fraud is a major global financial threat. This project builds a machine learning pipeline to automatically classify transactions as legitimate (0) or fraudulent (1) using historical transaction data. The core challenge is the extreme class imbalance — fraud cases account for only ~0.17% of all transactions.

Goal: Maximize fraud detection (Recall) while keeping false alarms (False Positives) low.

---

📊 Dataset

Property	Details
Source	Kaggle – Credit Card Fraud Detection
Total Records	284,807 transactions
Features	31 columns (Time, V1–V28, Amount, Class)
Target Column	Class → 0 = Legitimate, 1 = Fraud
Fraud Cases	492 (~0.17%)
Legitimate Cases	284,315 (~99.83%)
Missing Values	None
Duplicates	1,081 rows (removed during cleaning)

Note: Features V1–V28 are the result of PCA transformation to protect user confidentiality. Only Time and Amount are in their original form.

---

📁 Project Structure

credit-card-fraud-detection/
│
├── creditcard.csv                          # Raw dataset (download from Kaggle)
├── Credit_Card_Fraud_Detection.ipynb       # Main Jupyter Notebook
├── README.md                               # Project documentation
└── requirements.txt                        # Python dependencies

---

🛠️ Tech Stack

Category	Libraries / Tools
Language	Python 3.10+
Data Manipulation	pandas, numpy
Visualization	matplotlib, seaborn
Machine Learning	scikit-learn
Environment	Jupyter Notebook / Google Colab

---

🔄 Workflow

1. Import Libraries
       ↓
2. Load & Explore Dataset
   └── shape, head, tail, info, describe
       ↓
3. Data Cleaning
   ├── 3a. Handle Missing Values  → None found
   └── 3b. Remove Duplicate Rows → 1,081 removed
       ↓
4. Exploratory Data Analysis (EDA)
   ├── Q1: Class distribution (Fraud vs Legitimate %)
   ├── Q2: Fraud transaction amount distribution
   └── Q3: Amount comparison across both classes
       ↓
5. Model Development
   ├── 5a. Feature/Target split (X, y)
   ├── 5b. Train-Test split (80/20, stratified)
   └── 5c. Train Random Forest Classifier
       ↓
6. Model Evaluation
   ├── Classification Report
   ├── Confusion Matrix (numeric + heatmap)
   └── Feature Importance Plot
       ↓
7. Summary & Conclusions

---

📈 Key Results

Model: Random Forest Classifier (n_estimators=100, random_state=41)

Metric	Legitimate (0)	Fraud (1)
Precision	1.00	0.90
Recall	1.00	0.80
F1-Score	1.00	0.85
Support	56,671	75

Overall Accuracy: ~99.97%

Confusion Matrix

                  Predicted
                  Legit   Fraud
Actual  Legit  [ 56664     7  ]
        Fraud  [    15    60  ]

	Count
True Negatives (Legit correctly identified)	56,664
False Positives (Legit flagged as Fraud)	7
False Negatives (Fraud missed)	15
True Positives (Fraud correctly detected)	60

Out of 75 fraud cases in the test set, 60 were correctly detected and only 15 were missed.

---

📊 Visualizations

The notebook includes the following plots:

• Pie Chart — Class distribution (Fraud vs. Legitimate %)
• Histogram — Fraud transaction amount distribution
• Side-by-side Histograms — Amount comparison (Fraud vs. Legitimate)
• Confusion Matrix Heatmap — Visual breakdown of predictions
• Feature Importance Bar Chart — Top 15 most predictive features

---

▶️ How to Run

1. Clone or Download the Repository

git clone https://github.com/your-username/credit-card-fraud-detection.git
cd credit-card-fraud-detection

2. Install Dependencies

pip install -r requirements.txt

Or manually:

pip install pandas numpy matplotlib seaborn scikit-learn jupyter

3. Download the Dataset

Download creditcard.csv from Kaggle and place it in the project root directory.

4. Launch the Notebook

jupyter notebook Credit_Card_Fraud_Detection.ipynb

---

🚀 Possible Improvements

Technique	Purpose
SMOTE / Undersampling	Handle severe class imbalance
XGBoost / LightGBM	Potentially better performance
Hyperparameter Tuning	GridSearchCV or RandomizedSearchCV
ROC-AUC Curve	Better evaluation metric for imbalanced data
Precision-Recall Curve	Ideal metric for fraud detection
Cross-Validation	More robust model evaluation
Feature Scaling	Normalize Amount and Time columns

---

👨‍💻 Author

Rushikesh Sangamnere

Email: rushikeshsangamnere4561@gmail.com

Phone: +91 9096506345

---

📄 License

This project is for educational purposes. The dataset is publicly available on Kaggle under its own license terms.