CryptoPulse: Short-Term Cryptocurrency Forecasting with Dual-Prediction and Cross-Correlated Market Indicators

TL;DR

CryptoPulse tackles short-term cryptocurrency forecasting by integrating macro market environment signals, target-asset price dynamics, and market sentiment into a unified dual-prediction framework. A macro-conditioned branch and a price-dynamics branch produce two candidate next-day price predictions, which are fused through a sentiment-informed gate derived from LLM-based news analysis, with the fusion weight $\gamma$ computed as $\gamma = \zeta([\mathbf{x}_g^{emb}; \mathbf{s}^{emb}])$. The approach uses a mix of technical indicators, cross-cryptocurrency correlations, and few-shot prompting to新闻-driven regularization, achieving state-of-the-art results across multiple coins and market caps, and showing robustness in volatile regimes. The work has practical implications for automated trading and real-time crypto analytics by providing a scalable method to combine heterogeneous market signals into reliable short-horizon forecasts.

Abstract

Cryptocurrencies fluctuate in markets with high price volatility, posing significant 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 often overlook three critical factors influencing market dynamics: 1) the macro investing environment, reflected in major cryptocurrency fluctuations affecting collaborative investor behaviors; 2) overall market sentiment, heavily influenced by news impacting investor strategies; and 3) technical indicators, offering insights into overbought or oversold conditions, momentum, and market trends, which are crucial for short-term price movements. This paper proposes a dual prediction mechanism that forecasts the next day's closing price by incorporating macroeconomic fluctuations, technical indicators, and individual cryptocurrency price changes. Additionally, a novel refinement mechanism enhances predictions through market sentiment-based rescaling and fusion. Experiments demonstrate that the proposed model achieves state-of-the-art performance, consistently outperforming ten comparison methods.

Figures (10)

• Figure 1: Overview of CryptoPulse architecture for next-day closing price prediction.
• Figure 2: Deep Learning vs. traditional models on data without sentiment.
• Figure 3: Comparison of RNN-based models.
• Figure 4: MAE comparison between linear and transformer-based models.
• Figure 5: MSE comparison between linear and transformer-based models.