UPI Fraud Detection Using Machine Learning

Project Overview

With the rapid adoption of Unified Payments Interface (UPI) in India and other digital economies, the risk of fraudulent transactions has grown significantly. This project aims to build a machine learning model that can detect potentially fraudulent UPI transactions in real time, helping payment platforms and banks prevent financial losses and protect user data.

Tools & Technologies Used

Methodology / Process

Loading Datasets.
Data Preprocessing.
EDA and its parameters.
Data Correlation of selected columns.
Checking Imbalanced Data and downloading cleaned datasets.
Machine learning models : Decision Tree, Random Forest, Logistic Regression, Ada Boost, LGBM, ANN and Comparison.
Saving the model.

Loading Datasets

Data Preprocessing and converting date into datatime.

EDA and its parameters.

Plotting the fraud rate for each status using a bar chart.
Visualizing bar plot to check the fraud rate for each gateway.
Creating a bar plot to visualize the fraud rate across transaction channels and device operating systems.
Visualizing Fraud by city.
Visualizing Fraud by State.
Creating bar plots or box plots to visualize fraud rates and deviations.
KDE plots to visualize and compare the distribution of amount for both cases.
Compare fraud rates across different device operating systems.