Cryptocurrencymarketsarecharacterizedbyrapid fluctuations and non-stationary behavior, making accurate price prediction a persistent challenge. While numerous studies have examined Bitcoin using diverse machine learning models, researchonalternativecoinssuchasSolanaremainsscarcedespite its growing adoption and distinct market dynamics. This study develops a stacking ensemble model for Solana price prediction, integrating Support Vector Regression (SVR), Random Forest, and XGBoost as base learners. To better capture temporal patterns in price movement, the model incorporates lag features and exponential moving average (EMA) indicators derived from historical data. The dataset of Solana daily prices was prepro- cessedtoconstructthesefeaturesbeforetrainingthebaselearners andmeta-learner.Performancewasassessedusingmeanabsolute percentage error (MAPE), which enables direct comparison across models under volatile conditions. Experimental results show that the stacking ensemble achieved a MAPE of 3.27%, consistently outperforming the best single models. The results suggest that stacking ensembles with engineered lag and EMA features can provide more reliable predictive performance for Solana, offering a foundation for further research on forecasting alternative cryptocurrencies.