Introduction

Pump-and-dump schemes—coordinated buying to inflate prices, followed by dumps (sales) by original promoters—are illegal but pervasive in micro-cap stocks. Detecting them requires identifying coordinated buy/sell patterns. Temporal graph analysis, representing trader relationships and trade sequencing, can identify suspicious pump-and-dump patterns with machine learning, protecting retail investors.

Temporal Graph Construction

Model trades as a temporal graph: nodes are traders, edges represent trades (buyer-seller pairs), edges timestamped. Detect subgraphs with suspicious patterns: rapid circular trades (A sells to B, B sells to C, C sells to A), concentrated buys followed by concentrated sells from insiders, etc. Machine learning classifies graphs as normal vs suspicious.

Regulatory Application

Regulators can deploy graph-based detection systems to flag suspicious micro-cap pump schemes before significant losses. Investors avoiding flagged stocks protect themselves from losses.

Conclusion

Temporal graph analysis enables systematic detection of coordinated trading schemes, protecting markets from manipulation.