The paper presents an unsupervised pipeline for detecting suspicious trading activity. It applies K-Means++ to roughly one million financial transactions from 2012 to 2024, then uses market-practice heuristic thresholds to categorize flagged trades. The system marks 2.02% of transactions as suspicious: 51.10% as spoofing, 0.10% as pump and dump, 0.55% as insider trading, 1.43% as fake breakouts, and 46.83% as unclassified. However, the study has no ground-truth labels. Its Silhouette Score of 0.561 measures cluster separation, not the accuracy of fraud detection.
No heat snapshots are available in the last 24 hours.