<![CDATA[The rapid growth of cryptocurrency adoption has increased the risk of phishing attacks targeting blockchain transactions, particularly within the Ethereum ecosystem. This study aims to conduct a comparative analysis of machine learning approaches for detecting phishing transactions by addressing the challenge of highly imbalanced data. A supervised classification framework is developed using Random Forest and Light Gradient Boosting Machine algorithms, combined with the Synthetic Minority Over sampling Technique to enhance minority class representation. The research process involves data preprocessing, feature normalization, class balancing, model training, and performance evaluation using appropriate classification metrics. The findings indicate that ensemble based methods are effective in capturing complex transaction patterns and distinguishing legitimate from phishing activities. Random Forest demonstrates more robust and consistent performance compared to Light Gradient Boosting Machine, particularly in minimizing misclassification of fraudulent transactions. These results highlight the importance of data balancing techniques and algorithm selection in improving blockchain security. The study contributes to the development of reliable phishing detection models that can support safer cryptocurrency transaction environments.]]>
Copyrights © 2026
