Credit Card Fraud Detection using Machine Learning

A machine learning project focused on accurately identifying fraudulent credit card transactions from a highly imbalanced dataset. This project demonstrates a full data science workflow, from preprocessing and exploratory data analysis to model building, evaluation, and addressing class imbalance.

---

📖 Project Overview

Credit card fraud is a multi-billion dollar problem for the financial industry. This project tackles the challenge of building a predictive model that can precisely identify fraudulent transactions while minimizing false alarms. The core difficulty lies in the extreme class imbalance, where fraudulent transactions are a tiny fraction of all transactions.

Main Objective: To develop a robust machine learning model that achieves high recall (catching most fraud) without sacrificing too much precision (minimizing false positives).

📊 Dataset

Source: Credit Card Fraud Detection Dataset on Kaggle

• The dataset contains transactions made by European cardholders in September 2013.
• It comprises 284,807 transactions in total, with only 492 (0.172%) fraudulent ones.
• All features (V1 to V28) are numerical and are the result of a PCA transformation (Dimensionality Reduction) to protect confidentiality. The original features are not provided.
• Features Time and Amount are the only non-transformed features.
• Class is the target variable (1 = Fraud, 0 = Normal).

🛠️ Tech Stack & Libraries

• Programming Language: Python 3.8+
• Data Manipulation: Pandas, NumPy
• Data Visualization: Matplotlib, Seaborn
• Machine Learning: Scikit-learn, Imbalanced-learn
• Environment: Jupyter Notebook / Google Colab

🔍 Methodology & Approach

1. Data Preprocessing & Exploratory Data Analysis (EDA)

• Handled Missing Values: Checked for and confirmed no missing values.
• Analyzed Class Imbalance: Visualized the severe imbalance between fraudulent and non-fraudulent classes.
• Scaled Features: Standardized the Time and Amount features using StandardScaler to ensure all features were on a similar scale for the model.
• Train-Test Split: Performed a stratified split to maintain the same class distribution in both training and testing sets, ensuring a fair evaluation.

2. Dealing with Class Imbalance

To handle the extreme imbalance, two primary strategies were employed:

• Algorithm Choice: Used models like Random Forest that perform well on imbalanced data.
• Resampling Techniques: Experimented with Random Under-Sampling (reducing the majority class) to improve model performance on the minority class.

3. Model Building & Training

Trained and evaluated multiple classifiers to identify the best performer:

• Logistic Regression (Baseline)
• Random Forest Classifier
• XGBoost Classifier

The Random Forest model was selected for its strong performance and interpretability.

4. Model Evaluation

Models were evaluated based on metrics that are critical for imbalanced classification problems:

• Confusion Matrix: To visualize True/False Positives/Negatives.
• Precision: Ability to not label a legitimate transaction as fraudulent.
• Recall (Sensitivity): Ability to find all fraudulent transactions.
• F1-Score: Harmonic mean of Precision and Recall.
• Area Under the ROC Curve (AUC-ROC): Overall measure of model performance across all classification thresholds.

📈 Results & Key Findings

The final Random Forest model, tuned with hyperparameters, achieved the following results on the test set:

Metric	Score
Accuracy	99.95%
Precision	93.55%
Recall	82.05%
F1-Score	87.50%
AUC-ROC Score	0.981

• Key Insight: The model successfully identifies 82% of all fraudulent transactions while maintaining a high precision of 93.6%, meaning that over 9 out of 10 transactions flagged as fraud are actually fraudulent. This balance is crucial for a practical fraud detection system.