This project aims to develop a working model that can predict the price of Bitcoin on a 1-day timeframe. Various approaches were explored, including LSTM models, plain machine learning models, and transformers. The dataset includes Bitcoin price, price indicators, capitalization, and the greed/fear index.
The most promising model was the transformer, which utilized several days of previous price history to predict Bitcoin prices two weeks ahead.
-
Multiple Approaches:
- LSTM models
- Plain machine learning models
- Transformer model (most promising)
-
Dataset:
- Bitcoin price
- Price indicators
- Capitalization
- Greed/Fear index
-
Prediction:
- models predict Bitcoin price two weeks ahead