Fraud Detection Project (Using ML and Deployed using streamlit)

This project implements a machine learning-based fraud detection system for financial transactions. It includes a Jupyter notebook for exploratory data analysis (EDA) and model training, a Streamlit web application for interactive fraud predictions, and a dataset (https://www.kaggle.com/datasets/amanalisiddiqui/fraud-detection-dataset/data). The model leverages logistic regression with balanced class weights to address the imbalanced nature of fraud data, achieving high recall for detecting fraudulent transactions.

Table of Contents

• Project Overview
• Dataset
• Repository Structure
• Installation
• Usage
• Model Details
• Contributing
• License

Project Overview

The Fraud Detection Project aims to identify fraudulent financial transactions using machine learning. The Jupyter notebook (notebooks/Analysis_Model.ipynb) performs EDA, data preprocessing, and trains a logistic regression model to classify transactions as fraudulent or non-fraudulent. The Streamlit app (src/app.py) provides a user-friendly interface to input transaction details and receive real-time fraud predictions.

Key features:

• Analysis of transaction data with features like step, type, amount, and isFraud.
• Logistic regression model with balanced class weights to handle imbalanced data (0.13% fraud rate).
• Interactive web app for fraud prediction.

Dataset

The dataset (data/AIML Dataset.csv) contains 6,362,620 financial transaction records with the following key columns:

• step: Time step of the transaction.
• type: Transaction type (e.g., PAYMENT, TRANSFER, CASH_OUT).
• amount: Transaction amount.
• oldbalanceOrg, newbalanceOrig: Origin account balances before and after the transaction.
• oldbalanceDest, newbalanceDest: Destination account balances before and after the transaction.
• isFraud: Binary label (1 for fraudulent, 0 for non-fraudulent).

Note: If the dataset exceeds GitHub's 100 MB limit, it is hosted externally. Download it from [Insert External Link, e.g., Google Drive or Kaggle] and place it in the data/ folder.

Repository Structure

fraud-detection-project/
├── data/
│   └── AIML Dataset.csv
├── notebooks/
│   └── Analysis_Model.ipynb
├── src/
│   └── app.py
├── requirements.txt
├── README.md
├── LICENSE
└── .gitignore

• data/: Contains the dataset (AIML Dataset.csv).
• notebooks/: Includes the Jupyter notebook for EDA and model training.
• src/: Holds the Streamlit app code (app.py).
• requirements.txt: Lists Python dependencies.
• README.md: Project documentation.
• LICENSE: MIT License file.
• .gitignore: Excludes unnecessary files (e.g., virtual environments).

Deployed Streamlit App

The app is deployed on Streamlit Community Cloud. Access it at https://fraud-detection-app-taqhypmjn8savxewsib3yw.streamlit.app/.

Model Details

• Algorithm: Logistic Regression with class_weight='balanced' to handle the imbalanced dataset (0.13% fraud cases).
• Features: step, type (one-hot encoded), amount, oldbalanceOrg, newbalanceOrig, oldbalanceDest, newbalanceDest.
• Preprocessing:
  • Numerical features scaled using StandardScaler.
  • Categorical feature type encoded using OneHotEncoder.
• Performance (on test set):
  • Accuracy: 94.67%
  • Fraud class (isFraud=1):
    • Precision: 0.02 (low due to imbalance)
    • Recall: 0.94 (high, good at detecting fraud)
    • F1-Score: 0.04
  • See notebooks/Analysis_Model.ipynb for detailed metrics and confusion matrix.

Limitations:

• Low precision for fraud due to class imbalance.
• Future improvements: Use SMOTE for oversampling, test ensemble models (e.g., Random Forest, XGBoost), or tune hyperparameters.

📫 How to Reach Me

• 📧 Email:([email protected])
• 📧 Linkedin:(https://www.linkedin.com/in/kurra-srinivas-31727420b/)
• 📸 Instagram:(https://www.instagram.com/_srinivas.kurra/profilecard/?igsh=MWxuNnNpNXc2anhhMg==)