UPI FRAUD DETECTION USING MACHINE LEARNING ALGORITHMS

Abstract

Increase in UPI usage for online payments, Cases of fraud associated with it are also rising. Few steps 
involving UPI transaction process using a Hidden Markov Model (HMM). An HMM is initially trained for a 
cardholder. If a UPI transaction is not accepted by the trained HMM. It is considered to be fraudulent. People 
can use UPIs for online transactions as it provides an efficient and easy-to-use facility. With the increase in 
usage of UPIs, the capacity of UPI misuse has also enhanced. UPI frauds cause significant financial loses for 
both UPI holders and financial companies. In this project, the main aim is to detect such frauds, including the 
accessibility of public data, high-class imbalance data, the changes in fraud nature, and high rates of false 
alarm. The main focus has been to apply the recent development of machine learning algorithms for this 
purpose. We have created 5 Algorithms to detection the UPI Fraud and evaluated results based on that. Various 
modern techniques like artificial neural network. Different machine learning algorithms are compared, 
including Auto Encoder, Local Outlier Factor, Kmeans Clustering. This project uses various algorithms, and 
neural network which comprises of techniques for finding optimal solution for the problem and implicitly 
generating the result of the fraudulent transaction. This algorithm is a heuristic approach used to solve high 
complexity computational problems. The implementation of an efficient fraud detection system is imperative for 
all UPI issuing companies and their clients to minimize their losses.