| dc.description.abstract | The cryptocurrency market, particularly Ethereum, is known for its high volatility and the complex factors that influence its price. This makes Ethereum price prediction a significant challenge, especially for investment decision-making. This study aims to evaluate the capability of the Random Forest algorithm in predicting Ethereum prices and assess the impact of Random Search optimization on the model's performance. The research uses five years of historical Ethereum data with features including open, high, low, close, volume, and percentage change. The results show that the non-optimized Random Forest model achieved a Mean Absolute Percentage Error (MAPE) of 2.82% and a Root Mean Square Error (RMSE) of 101 USD for one-day-ahead predictions. After applying Random Search optimization, the model performance improved, with MAPE arround 2.60% and 96.5 USD of RMSE. However, the prediction accuracy tends to decrease over longer prediction horizons, reaching a MAPE of up to 10.2% by the tenth day, indicating that the model is only good for short-term forecasting. In conclusion, Random Search optimization enhances the accuracy of the Random Forest algorithm, although the model still struggles with long-term predictions. Another limitation lies in the use of limited features, relying solely on historical price values without considering external factors such as multi-week trends or macroeconomic conditions that may affect Ethereum's price. Future studies are recommended to incorporate additional features to improve model performance. | en_US |
