Fake UPI Transaction Detection Using Random Forest Algorithm

Authors:

¹Nadha Gopal, ²Naren, ³Imtha, ⁴Dr.Anand ⁵Dr. T. Kumanan, ⁶Dr. M. Nisha.

1,2,3 Students, Department of CSE

4,6 Assistant Professor, Department of CSE 5 Professor, Department of CSE

Dr.M.G.R Educational and Research Institute, Maduravoyal, Chennai 95, Tamilnadu, India

Abstract- Rapid online transactions have become the norm in business, and this is possible because UPI is fast and reliable in managing money. Nevertheless, trouble is increasing with the number of online transactions, fraudulent messages mislead people, identity thieves steal accounts, rogue requests appear out of thin air, and accounts are compromised without people’s knowledge. When individuals become victims, both wallets and trusts are hit by the losses as they continue to accrue.

The urge to detect frauds in their early stages keeps the financial scholars in the thick of transnational security initiatives. The traditional tools of detecting frauds cannot keep up with emerging types of frauds. It is so as they do the work to check set rules only. Rather than following the rigid designs, smarter ways have been discovered. These systems are able to learn by past data to identify subtle variations in data that are not perceived by others. Given huge quantities of information, they discover what orthodox systems miss. It is also more accurate since the adjustments are made automatically.

The current report presents a solution with the help of Random Forest to identify fake UPI deals. It analyzes the different details such as money involved, the timing, the clues of the device, position, frequency of transactions made as well as user behavior as opposed to depending on few factors. Prior to any analysis taking place, such steps as addressing gaps in data, transforming raw variables into useful formats, rescaling, and uneven counts of categories precede. It is only after these steps that the patterns start appearing clearly in the set up of the model. The Random Forest approach combines many decision trees as opposed to having a single tree model. The category improves reliability and also reduces the probability of having overly influenced results due to minor data patterns.

Findings indicate that the new technique is able to identify fake transactions with high accuracy, which have high marks in accuracy, recall, precision and F1-score. Random Forest is more effective than single-rules systems at noise resistance due to the fact that it makes decisions based on a combination of multiple-rules. Such a model is suitable within an app that utilizes UPI-based payment and serves to notice scams in the process and improve the level of overall safety.

Keywords— UPI Fraud Detection, Random Forest, Machine Learning, Financial Security, Classification, Digital Payments