Recently, Solana has had many opportunities for secondary trading. Compared to PVP, I am better at secondary trading and have developed the idea of using feature engineering and machine learning to predict the appropriate timing for token buying. Use Grok DeeperSearch to collect relevant references, provide a solution and development requirements outline, use Gemini Pro 2.5 to write initial code based on the requirements outline, and use Claude 3.7 to debug the code. This tweet records the development process of the prediction model