This project provides a deep learning model for predicting Bitcoin prices up to 9 minutes into the future using a GRU recurrent neural network. It is targeted at individuals interested in algorithmic trading or exploring time-series forecasting with neural networks. The primary benefit is a readily available, albeit simple, deep learning model for short-term cryptocurrency price prediction.

How It Works

The project utilizes a Gated Recurrent Unit (GRU) neural network, a type of recurrent neural network well-suited for sequential data like time-series. The GRU processes historical price data to learn patterns and predict future values. The model is trained and run using Python scripts, with data normalization being a key step in both processes.

Quick Start & Requirements

• Install: Clone the repository.
• Requirements: Python 3, Keras, NumPy, Matplotlib.
• Usage:
  • Training: python3 network.py -train dataset_path -iterations number_of_training_iterations
  • Finetuning: python3 network.py -train dataset_path -iterations number_of_training_iterations -finetune base_model_path
  • Running: python3 network.py -run dataset_path -model model_path
• Data: Download data via grabber.py.

Highlighted Details

• Predicts Bitcoin price for the next 9 minutes.
• Employs a GRU recurrent neural network.
• Includes options for training from scratch and finetuning existing models.
• Generates a chart.png visualizing predicted vs. real values.

Maintenance & Community

No information on maintenance, contributors, or community channels is provided in the README.

Licensing & Compatibility

The README does not specify a license.

Limitations & Caveats

The project's effectiveness is limited to very short-term predictions (9 minutes) and relies on a single GRU model without advanced ensemble methods or hyperparameter tuning. The README does not mention performance benchmarks or validation metrics.