Abstract:
"Identification of Anomalies in Online Transactions [Anomaly detection] that can eliminate Risk such as Financial Frauds,
Illegal Money Transaction and Anti-Money Laundering is very important. With the cryptocurrency market moving quickly, there is a rise in the need for preventing fraud especially with Ethereum (an open-source blockchain platform that enables developers to build and execute smart contracts). In 2020, Cryptocurrency frauds in the United States reached over 80,000 cases just that year and other countries such as Australia and UK being no different from having same challenges with numbers of his sort. This study proposed a hybrid analysis for detecting fraudulent transactions on the Ethereum network using machine learning and deep learning techniques. In particular, it utilizes Decision Trees, Neural Networks, Random Forest classifiers and SVM as well as Deep Convolutional Neural Network (CNN) models for echocardiogram classification with a focus on the Random Forest classifier. Following strict parametric evaluation and statistical analyses using Fisher's F-Test (p-value ¡ 0.001)) the Random Forest outperformed all other classifiers with an accuracy of 95.56%. This makes it effective in reducing the overfitting problem related to decision trees, and subsequently improving classification accuracy. Our results emphasize the need for extracting features from complex smart contracts and identifying anomalous transaction. The proposed model can serve as a secure way of validating cryptocurrency transactions, especially within the Ethereum ecosystem, which signals sustained and increased consumer adoption."
