As digital assets like cryptocurrencies have been more integrated into global financial operations, the financial markets have seen a dramatic shift in the past few years. Crypto markets are notoriously unpredictable; therefore, accurate prediction models are crucial for gauging future price swings. To optimize hyperparameters and improve prediction accuracy, we present a GA-CNN-LSTM model, a hybrid deep learning approach. To draw useful conclusions from the past performance of four prominent cryptocurrencies—Binance (BNB), Solana (SOL), Cardano, and Litecoin (LITE)—this model employs a combination of the Genetic Algorithm (GA) and CNN-LSTM. We conducted a full empirical investigation to see how well it worked. Also, for comparison's sake, we built and used several other models for the same job: CNN, LSTM, GA-CNN, GA-LSTM, and CNN-LSTM. The accuracy of each model was evaluated by calculating its Mean Absolute Error (MAE) and Mean Squared Error (MSE). The GA-CNN-LSTM model avoids overfitting and delivers better performance using GA-based tweaking. It also outperforms other models when using various cryptocurrency datasets.