Zejnilovic, LeidGiesbrecht, Markus2025-08-042025-08-042025-01-242024-12-17http://hdl.handle.net/10362/185999XRP is among the top 5 most prevalent cryptocurrencies as of 2024. However, the pseudonymous nature of XRP Ledger transactions complicates the understanding of economic activities and regulatory oversight of fraud within the network. This study uses descriptive statistics to develop heuristics for categorizing distinct types of economic agent activities and leverages unsupervised machine learning to segment agents into clusters; successfully distinguishing accounts such as decentralized exchanges, gambling sites, and NFT-related entities. Additionally, supervised fraud detection models are trained with an off-chain dataset of accounts involved in spam, token theft, and Ponzi schemes, achieving fraud detection accuracies of over 90%.engBehavioral analysisBlockchainFraud detectionHeuristicsLedgerlyticsMachine learningRippleSegmentationSupervisedUnsupervisedXRP Ledgermaster thesis203962877