Abstract
Cryptocurrencies fluctuate in markets with high price volatility, posingsignificant challenges for investors. To aid in informed decision-making,systems predicting cryptocurrency market movements have been developed,typically focusing on historical patterns. However, these methods oftenoverlook three critical factors influencing market dynamics: 1) the macroinvesting environment, reflected in major cryptocurrency fluctuations affectingcollaborative investor behaviors; 2) overall market sentiment, heavilyinfluenced by news impacting investor strategies; and 3) technical indicators,offering insights into overbought or oversold conditions, momentum, and markettrends, which are crucial for short-term price movements. This paper proposes adual prediction mechanism that forecasts the next day's closing price byincorporating macroeconomic fluctuations, technical indicators, and individualcryptocurrency price changes. Additionally, a novel refinement mechanismenhances predictions through market sentiment-based rescaling and fusion.Experiments demonstrate that the proposed model achieves state-of-the-artperformance, consistently outperforming ten comparison methods.