Enhancing Digital Payment Security with AI-Driven Biometric Authentication and Big Data Analytics: A Scalable Approach for Secure Remote Commerce

Abstract

This conducted research investigates the application of machine learning models to enhance fraud detection in financial transactions. The research work at hand uses a Gradient Boosting Classifier, Random Forest Classifier, and Support Vector Machine models with the main aim of improving fraud detection. The set contains records of fraudulent and non-fraudulent transactions relating to the transaction amount, means of payment, and biometric data. Pre-processing steps may include changing categorical variables into numeric, mapping the target variable into binary values, and scaling features. Each model must be trained and scored on metrics about accuracy, precision, recall, and F1 score. A Gradient Boosting Classifier is an algorithm where performance has been polished iteratively by updating the weights over the misclassified instances. The Random Forest Classifier improves the decision of weak classifiers by aggregation. SVM finds the best hyperplane to separate fraudulent and non-fraudulent transactions. The performances through these models are compared to searching for the best method. The results reveal strengths and weaknesses of each model and draw out insights on how effective these are in real fraud detection cases.