An Explainable Hybrid CNN–BiLSTM–Attention Model for Solana Price Forecasting Using Technical Indicators in Blockchain-Based Metaverse Economies
International Journal Research on Metaverse · 02.11.2026 · 3 (4) · 331-353
DOI: 10.47738/ijrm.v3i4.90 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal Research on Metaverse, online erschienen am 02.11.2026. Verfasst von Bentar Priyopradono.
Abstract (Originalsprache)
The rapid growth of blockchain-based metaverse ecosystems has increased the importance of accurate cryptocurrency price forecasting to support investment decisions, digital asset management, and decentralized application development. As one of the leading blockchain platforms powering non-fungible tokens (NFTs), decentralized finance (DeFi), GameFi, and metaverse applications, Solana exhibits high price volatility that poses significant challenges for market participants. Although deep learning models have demonstrated promising forecasting performance, many existing approaches remain difficult to interpret due to their black-box nature. This study proposes an explainable Hybrid Convolutional Neural Network, Bidirectional Long Short-Term Memory, and Attention (CNN–BiLSTM–Attention) framework for one-day-ahead Solana price forecasting using historical market data and technical indicators. The proposed framework integrates convolutional feature extraction, bidirectional temporal dependency learning, and an attention mechanism to improve forecasting performance, while SHapley Additive exPlanations (SHAP) and permutation feature importance are employed to enhance model interpretability. Experimental evaluation was conducted using 999 daily historical Solana price records, resulting in 599 samples after feature engineering and preprocessing. The proposed model achieved an MAE of 0.7470, an RMSE of 0.9232, a MAPE of 3.43%, and an R² of 0.6929, demonstrating satisfactory forecasting performance for short-term price prediction. Explainability analysis identified RSI-14, MACD, SMA-5, Open Price, and lagged price variables as the most influential features contributing to model predictions. The findings indicate that the proposed framework not only provides reliable forecasting performance but also improves model transparency, thereby supporting more informed decision-making within blockchain-based metaverse economies. Future work will investigate multimodal data integration and more advanced attention mechanisms to further improve forecasting accuracy and robustness.
Zitieren
APA
Priyopradono, B. (2026). An Explainable Hybrid CNN–BiLSTM–Attention Model for Solana Price Forecasting Using Technical Indicators in Blockchain-Based Metaverse Economies. International Journal Research on Metaverse, 3(4), 331-353. https://doi.org/10.47738/ijrm.v3i4.90
Vancouver
Priyopradono B. An Explainable Hybrid CNN–BiLSTM–Attention Model for Solana Price Forecasting Using Technical Indicators in Blockchain-Based Metaverse Economies. International Journal Research on Metaverse. 2026;3(4):331-353. doi:10.47738/ijrm.v3i4.90
BibTeX
@article{Priyopradono2026,
author = {Priyopradono, Bentar},
title = {An Explainable Hybrid CNN–BiLSTM–Attention Model for Solana Price Forecasting Using Technical Indicators in Blockchain-Based Metaverse Economies},
journal = {International Journal Research on Metaverse},
year = {2026},
volume = {3},
number = {4},
pages = {331--353},
doi = {10.47738/ijrm.v3i4.90}
}
EUVIMED