Author : Dhruv Patel

Fraudulent activity detection within blockchain networks has become a critical concern due to the widespread adoption of decentralized technologies in financial and digital systems. The paper introduces a system that uses Blockchain and Machine Learning (ML)to strengthen the security of banks. Employing the services of the Ethereum blockchain dataset, the model applies a comprehensive methodology involving data preprocessing, feature engineering, Z-score normalization, and stratified data splitting. Genetic Algorithm-optimized Support Vector Machine (GA-SVM) and Artificial Neural Network (ANN) are constructed and tested, and their results are then compared with those from Generalized Autoregressive Conditional Heteroskedasticity (GARCH) and Convolutional Neural Network (CNN) models. Metrics of accuracy by using Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) as measures. It was found that the GA-SVM model achieved the best results compared to other models, with MAE at 0.1032 and MAPE at 4.6938 on test data, which confirms its usefulness in real-time fraud detection. When the model connects with smart contracts, it helps prevent fraudulent activities and supports both transparency and good operations in blockchain-based finance.

Keywords: Blockchain, fraud detection, banking security, GA-SVM, ANN, Ethereum dataset, machine learning, MAE, MAPE, classification.