Deep Learning Approaches to Cryptocurrency Price Prediction

Leverage existed CryptoMamba structure with sentiment score and compare with other algorithms

Repository Includes:

• Implementation of CryptoMamba, LSTM, iTransformer, and Temporal Convolutional Network (TCN)
• Code for data preprocessing, model training, evaluation metrics

Quick Start (modifications needed)

Configuration

• Data Config: configs/data_configs/mode_1.yaml - defines data source, time intervals, and splits
• Training Configs:
  • configs/training/cmamba_nv.yaml - without volume
  • configs/training/cmamba_v.yaml - with volume
• Model Configs:
  • configs/models/CryptoMamba/v1.yaml - 5 features, no normalization
  • configs/models/CryptoMamba/v2.yaml - 6 features, no normalization

Data Format

Your raw data CSV should contain:

• Timestamp (or Date column)
• Open, High, Low, Close prices
• Volume (optional, for volume variant)

Model Architecture

CryptoMamba uses:

• Mamba SSM blocks with selective state updates
• Residual connections and optional MLP branches
• Configurable hidden dimensions and layer density
• Window-based processing (default: 14 days)

Outputs

• Logs: logs/ (TensorBoard)
• Checkpoints: checkpoints/ (best models)
• Results: Results/<name>/<config>/ (evaluation plots)
• Predictions: Predictions/<config>/<date>.txt (next-day forecasts)

Requirements

• PyTorch
• PyTorch Lightning
• mamba-ssm[causal-conv1d]
• pandas, numpy, matplotlib, seaborn
• See requirements.txt for full list

Citation

@article{Sepehri2025CryptoMamba,
    title={CryptoMamba: Leveraging State Space Models for Accurate Bitcoin Price Prediction}, 
    author={Mohammad Shahab Sepehri and Asal Mehradfar and Mahdi Soltanolkotabi and Salman Avestimehr},
    year={2025},
    url={https://arxiv.org/abs/2501.01010}
}