A collection of tools and experiments around financial data modeling. All the useful stuff is in the stock-prediction
folder. Notable features:
- Expression language for defining and transforming Signals (time-series arrays). Signals can be pulled from APIs, databases, files, etc. For example
rsi<window=14>(candles<interval=5min>[close])
will give you the 14-period RSI of the 5-minute candles' close time. - Simple API for training, evaluating, and (de)serializing models.
- Uses polars for many core features (still working on the transition from pandas).
This is not the latest version of the code - contact me if this is useful, and I will update it.