Enhancing Financial Market Prediction: A Comparative Analysis of RNN and LSTM Models for Cryptocurrency Price Forecasting

Advances in deep learning models such asRecur-rent Neural Networks (RNNs) and Long Short-TermMemory (LSTM) networks have shown promise in forecastingfinancial markets, yet their application to cryptocurrencyremains un-derexplored. A detailed evaluation of both RNNand LSTM models is conducted, with a focus on Ethereumprice prediction using ten years of historical data. The studycompares model performance in capturing the intricate, non-linear dynamics of highly volatile cryptocurrency markets.Results indicate that LSTMs outperform RNNs in learninglong-term dependencies, but both models face limitations innoisy, non-stationary market conditions. A hybrid approachintegrating traditional technical indicators—Moving Averages,Relative Strength Index (RSI), and Bollinger Bands—demonstrates improved prediction accuracy. The researchhighlights the potential for real-time learning and dynamicfeature selection in enhancing model adaptability to suddenmarket changes, offering new insights into the capabilities andchallenges of applying deep learning to cryptocurrency trading.