Artificial intelligence-based model – a key technique in digital economy development for cryptocurrency price prediction

This research examines the application of Long Short-Term Memory (LSTM) neural networks for predicting cryptocurrency prices, with a focus on Bitcoin (BTC) and Ethereum (ETH), the two dominant digital assets with the highest market capitalization. The study addresses the critical challenge of accurately forecasting cryptocurrency price movements in highly volatile markets, which is essential for informed investment decision-making in the digital economy. The methodology employs LSTM models trained on historical closing price data from 2014 to 2024 for Bitcoin and from 2017 to 2024 for Ethereum, utilizing an 80:20 training-to-testing ratio. Results demonstrate exceptional predictive accuracy with R² values of 99.08% for Bitcoin and 97.44% for Ethereum, while MAPE values remained low at 1.8% and 1.9%, respectively. The study concludes that LSTM models effectively capture complex patterns in cryptocurrency price movements, providing reliable short-term forecasting capabilities and contributing valuable insights to the intersection of artificial intelligence and digital economy development.