Improving Cryptocurrency Pump-and-Dump Detection through Ensemble-Based Models and Synthetic Oversampling Techniques

Paper: arXiv 2510.00836

Authors: Jieun Yu, Minjung Park, Sangmi Chai

Abstract

This study aims to detect pump and dump (P&D) manipulation in cryptocurrency markets, where the scarcity of such events causes severe class imbalance and hinders accurate detection. To address this issue, the Synthetic Minority Oversampling Technique (SMOTE) was applied, and advanced ensemble learning models were evaluated to distinguish manipulative trading behavior from normal market activity. The experimental results show that applying SMOTE greatly enhanced the ability of all models to detect P&D events by increasing recall and improving the overall balance between precision and recall. In particular, XGBoost and LightGBM achieved high recall rates (94.87% and 93.59%, respectively) with strong F1-scores and demonstrated fast computational performance, making them suitable for near real time surveillance. These findings indicate that integrating data balancing techniques with ensemble methods significantly improves the early detection of manipulative activities, contributing to a fairer, more transparent, and more stable cryptocurrency market.

Complexity vs Empirical Score

• Math Complexity: 2.5/10
• Empirical Rigor: 6.5/10
• Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper applies well-known machine learning models (XGBoost, LightGBM) and standard oversampling techniques (SMOTE) to detect cryptocurrency pump-and-dump schemes. While it lacks deep theoretical math or highly novel algorithms, it addresses a practical market surveillance problem with clear methodology and promising empirical results.

Research Flowchart

  flowchart TD
    A[Goal: Detect Crypto Pump-and-Dump Events] --> B[Input: Imbalanced Crypto Market Trading Data]
    B --> C[Data Balancing: Apply SMOTE Oversampling]
    C --> D[Model Training: Train Ensemble Models<br/>XGBoost, LightGBM, etc.]
    D --> E[Evaluation: Measure Recall, Precision & F1-Score]
    E --> F[Outcomes: High Recall XGBoost 94.87%, LightGBM 93.59%<br/>Fast performance for real-time market surveillance]