Forecasting Volatile Cryptocurrency Prices: An LSTM Approach Applied to Solana with Advanced Time Series Features

Main Article Content

👤 Chendri Irawan Satrio Nugroho
🏢 Department of Information Technology, Institut teknologi Tangerang Selatan, Indonesia
👤 Erland Inkiriwang
🏢 Department of Informatics, Institut teknologi Tangerang Selatan, Indonesia

This study aims to develop a deep learning model based on Long Short-Term Memory (LSTM) to forecast Solana cryptocurrency prices using historical time series data. The dataset is enriched with multiple technical indicators, including moving averages, volatility measures, and log return transformation to improve data representation. The model is trained using a sequence length of 30-time steps and evaluated using several performance metrics, namely Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²). The experimental results show that the proposed model achieves a MAPE value below 5%, indicating high prediction accuracy, and an R² value that reflects a strong ability to capture the underlying patterns in the data. The model is able to effectively follow overall price trends and short-term fluctuations, although its performance decreases during periods of high volatility, where sudden price changes occur. These findings suggest that the LSTM model is a suitable approach for modeling non-linear and sequential patterns in cryptocurrency markets, while highlighting the need for further improvements to enhance performance under extreme market conditions.

Irawan Satrio Nugroho, C., & Inkiriwang, E. (2026). Forecasting Volatile Cryptocurrency Prices: An LSTM Approach Applied to Solana with Advanced Time Series Features. Fintech Innovation Journal, 2(4), 294–308. https://doi.org/10.63913/ftij.v2i4.42

Article Details

Section
Articles

Similar Articles

<< < 1 2 3 

You may also start an advanced similarity search for this article.