Credit Card Fraud Detection System

Demo

Click on the thumbnail to watch the video!

Project Synopsis: Credit card fraud is a significant problem in the financial industry, resulting in billions of dollars in losses annually. Traditional fraud detection systems often rely on manual reviews and simple rule-based systems, which can be time-consuming and ineffective. This project aims to develop a more efficient and accurate system using advanced data structures and algorithms The project will focus on developing a robust and efficient system that can analyze datasets of credit card transactions and detect fraudulent activities. The system will utilize a combination of data structures, including hashmap,doubly linked lists, and binary search trees, to store and process transaction data.

Data Structures Employed:

• Hash-map: Used to store statistical data such as mean, standard deviation and to maintain references to linked lists and BST nodes for multiple users.
• Binary Search Tree (BST): Employed for efficient storage and retrieval of transaction amounts and statuses, allowing for quick look up based on dates.
• Linked List: Utilised to track transactions based on timestamp and location, supporting chronological ordering and traversal of transaction history.

Data Flow and Integration: At program execution, user enters his/her credit card number and password, and after the login is successful, transaction data is loaded into data structures. A menu driven program to search for transaction (based on date and location), showing flagged transactions and predicting the possibility for fraud.

Fraud Detection:

• if multiple failed transaction occur in a short span of time, they are a sign of potential fraud and will be flagged as the same. Monitor the time intervals between consecutive transactions. Flag any transactions with unusually short time intervals as potential fraud.
• if the location of the current transaction is significantly different from the previous transaction’s location.
• if neither time nor location anomalies are detected, apply a z-score-based statistical method to compare the current transaction amount against the user’s spending pattern. If the z-score exceeds a predefined threshold, flag it as a potential fraud.

Algorithms:

1. Naive Bayes: Predicts whether a transaction is fraudulent based on historical data and probabilities.
2. Z-Score Calculation: Flags transactions significantly deviating from a user’s average spending.
3. Time Interval Analysis: Detects rapid consecutive transactions, which could indicate suspicious activity.
4. Location Anomaly Check: Flags geographically inconsistent transactions in short time intervals.

References used :

• Krish Naik's Naive Bayes Tutorial
• A Credit card fraud detection using Naïve Bayes and Adaboost Research Paper
• Naïve Bayes classifiers - Article by IBM

Datasets Employed

• Generating Credit Card Details for users
• Password Generator
• Daily Transaction Dataset from Kaggle
• Generate Credit Card Transaction Data

Contributors

• @Shloka-Shinde
• @vaish-navi24
• @Sanket-Sonawane-cpp