CREDIT CARD FRAUD DETECTION

Fraud Detection System for Financial Transactions

Problem Statement

Banks and financial institutions face significant financial losses due to fraudulent transactions. With the increasing volume of digital payments, detecting fraud accurately and in real time has become critical.

We were provided with transactional data containing features such as transaction time, amount, merchant details, location, and customer information.

Goal:

To build a machine learning model capable of accurately identifying fraudulent transactions while minimizing false positives to maintain customer trust and experience.

Project Objective

Develop a robust fraud detection model.

Automatically flag suspicious transactions.

Reduce financial losses caused by fraud.

Maintain customer satisfaction by avoiding unnecessary transaction blocks.

Models Implemented

To determine the most effective model, multiple algorithms were trained and evaluated:

Logistic Regression

Logistic Regression with 10-Fold Cross-Validation

Random Forest

Random Forest with 10-Fold Cross-Validation

Decision Tree

Decision Tree with 5-Fold Cross-Validation

Why We Chose Random Forest

Random Forest was selected because:

It performs strongly on tabular datasets containing both categorical and numerical features.

It handles imbalanced datasets better than simpler models.

It reduces overfitting by combining multiple decision trees.

It provides stable and consistent performance across validation folds.

Model Evaluation

We used K-Fold Cross-Validation (k = 10) to ensure robust evaluation and avoid overfitting.

Best Performance Results (k = 10):

Average F1 Score (Normal): 0.999

Average F1 Score (Abnormal): 0.999

Average Recall (Normal): 0.998

Average Recall (Abnormal): 0.999

Average Precision (Normal): 0.999

Average Precision (Abnormal): 0.998

Average Accuracy: 0.999

Key Results

The model automatically flags suspicious transactions in real time.

Achieved extremely high precision and recall, minimizing both false positives and false negatives.

Significantly reduces financial losses.

Preserves customer trust by maintaining a smooth transaction experience.