Real-Time Fraud Detection and Securing Customer Transactions in E-Commerce Using Gradient Boosting Machines

E-commerce platforms must secure transactions due to the many frauds attempts and dangers that harm customers and enterprises. Traditional rule-based fraud detection systems can’t adapt to changing fraud types. However, ML models may dynamically scale up or down to detect threats in real time. These programs search mounds of data for anomalies that could suggest fraud to protect consumer financial transactions. This study presents OpLiGBM, an intelligent fraud detection system, to secure client financial transactions. Preprocessing included Z-scores, categorical variable encoding, and imputation for missing values to improve model performance. PCA detected outliers, and Bayesian-based hyperparameter optimisation optimised LightGBM settings. In fraud detection dataset studies, OpLiGBM exceeded precision 94.21%, accuracy 96.73%, area under the curve 97.48%, and F1-score 96.87%. The findings show how well the proposed technique detects fraud. Real-time fraud detection protects financial transactions and reduces losses. This method is accurate and efficient, making it a valuable tool for fraud detection systems. Future research can improve detection by adding data sources and using ML.