AI Predicts Solana Memecoin Rug Pulls Early Using Machine Learning

The rapid growth of memecoins on blockchain platforms like Solana has unfortunately led to an increase in fraudulent activities, particularly "rug pulls" where developers abandon a project after raising funds. Unlike Ethereum, where smart contract vulnerabilities are often exploited, Solana rug pulls primarily involve manipulating liquidity and social sentiment. This study addresses the urgent need for early detection by creating a large dataset of 6.4 million Solana tokens over seven months. The research found that a significant majority of these memecoins exhibit rug pull characteristics within an hour of launch. Despite the lack of code-level features, classic machine learning models, specifically Gradient Boosting (XGBoost), proved effective in identifying potential rug pulls using only the initial five minutes of trading data. The study also explored cross-platform generalization between PumpFun and Raydium, demonstrating that combining data from multiple sources enhances detection reliability and mitigates domain shift. This work provides a practical framework for safeguarding investors and deepens the understanding of decentralized finance fraud on high-throughput blockchains. It highlights the potential for AI to act as an early warning system in volatile crypto markets.