How amingolnari/demo-btcusd-pricepred-probabilisticdl-transferlearning works
Project type and purpose
• This is not a multi-service system but a single Jupyter Notebook demo for cryptocurrency price prediction.
• Purpose: showcase probabilistic deep learning (GRU/LSTM variants via TensorFlow Probability) and transfer learning on Bitcoin price data, then apply the best model to other cryptocurrencies.
File structure
• Top level contains only:
• Demo_BTCUSD_PricePred_5min_TensorFlowProb.ipynb
• README.md
• No separate “src”, “backend” or “frontend” directories—everything runs in one notebook.
• No deployment or service orchestration files; this is an experimental research demo.
README for additional insights
• Workflow stages clearly described: data download (Yahoo Finance), preprocessing, model training (multiple architectures), transfer learning, evaluation, and visualization.
• Dependencies: TensorFlow 2.15.1, Keras 2.15.0, TensorFlow Probability 0.23.0, yfinance.
• Models: various probabilistic and simple GRU/LSTM variants, bidirectional and time-distributed variants.