The rapid growth of cryptocurrencies has enabled the use of decentralized networks for complex money laundering activities and the creation of new financial services. In general, the functionality of conventional rule-based monitoring systems is surpassed by contemporary money laundering techniques on blockchain networks. In order to identify anomalous or dubious patterns for widespread application, machine learning can evaluate extensive transaction data. In order to detect behavioral indicators of illicit activity, the proposed methodology implements both supervised and unsupervised models. In order to facilitate identification, it is necessary to acquire critical data, including network architecture, transaction frequency, wallet interactions, and temporal trends. In order to improve the model’s stability and accuracy, we implement feature engineering and dimensionality reduction strategies. The algorithm is instructed by annotated datasets that include both legal and illicit transactions. Experimental assessment exhibits superior accuracy and improved information retention when contrasted with conventional detection methods. This method allows for the rapid identification of wallets and transaction patterns that are at a high risk.