Fraud Detection EDA using Python

Project Overview

This project uses Python-based exploratory data analysis (EDA) to investigate transaction patterns and identify potential fraud-risk signals. The analysis focuses on transaction behaviour, merchant activity, client-level risk indicators and seasonal patterns.

Business Problem

Fraud and unusual transaction behaviour can create financial losses, operational workload and customer trust issues. The aim of this project is to use EDA to highlight where a risk team should investigate further, rather than to claim a fully deployed fraud prediction system.

Tools Used

• Python
• Pandas
• NumPy
• Matplotlib
• Seaborn
• Jupyter Notebook

Repository Structure

fraud-detection-eda-python/
├── README.md
├── PROJECT_SUMMARY.md
├── requirements.txt
├── data/
│   └── README.md
├── docs/
│   └── methodology.md
├── notebooks/
│   └── fraud_detection_eda.ipynb
└── outputs/
    └── charts/

Analysis Covered

• Data understanding and cleaning
• Transaction trend analysis over time
• High-value transaction analysis
• Merchant-level transaction pattern review
• Transaction type analysis
• Top client analysis
• Unique card usage by client
• High-risk client categorisation
• Monthly spending pattern analysis for high-risk clients

Key Business Insights

• High-value transaction activity can vary meaningfully by month, making seasonal monitoring useful.
• Certain merchants and clients show substantially higher transaction volumes, making them useful targets for deeper risk review.
• Online transaction behaviour can be used as a screening feature for potential high-risk clients.
• Fraud monitoring should combine transaction value, channel behaviour, merchant concentration and client-level patterns.

How to Run

Install the required libraries:

pip install -r requirements.txt

Open the notebook:

jupyter notebook notebooks/fraud_detection_eda.ipynb

Dataset Note

This project uses the Kaggle dataset Financial Transactions Dataset: Analytics by ComputingVictor.

Dataset link: https://www.kaggle.com/datasets/computingvictor/transactions-fraud-datasets

The full transactions_data.csv file is not included in this repository because of file-size limitations. To reproduce the analysis, download the dataset from Kaggle and place transactions_data.csv inside the data/ folder.

Expected path:

data/transactions_data.csv

## Recruiter Summary
This project demonstrates Python EDA, data cleaning, risk-focused analysis and business interpretation for fraud monitoring use cases.