Abstract

Prior to machine learning, businesses would employ a rule-based strategy to identify fraud by examining recurring and obvious indicators. Countless fraud scenarios are executed by pure rule-based algorithms, which are personally crafted by an individual. The goal of the paper is to create precise deep learning and machine learning models for the Green Finance fraud detection. Fraud in real-time transactions cannot be detected by conventional rule-based methods. The study tackles the problem of unbalanced data by using the PaySim dataset, which replicates mobile transactions. The performance of a number of algorithms is assessed, including Random Forests, Recurrent Neural Networks, and K-Nearest Neighbors. To uncover hidden patterns in user transactions, the application of long short-term memory models and artificial neural networks is investigated. The study talks about difficulties including data cleaning and tweaking hyperparameters. The results support the development of more precise and effective fraud detection systems, which helps Green Fiannce lower losses and preserve consumer confidence.