Bulbea

Bulbea is an open-source Python library with 2,300+ stars on GitHub that applies deep learning to stock market prediction and modelling, wrapping data loading, preprocessing, model training and evaluation into one small toolkit.

Overview

Built on TensorFlow, Bulbea treats a stock as a first-class object: you create a Share for a ticker such as GOOG from Yahoo Finance data, and the library handles the Open, High, Low, Close and Volume series for you. From there it provides the pieces a quantitative experiment needs — splitting a dataset into training and test sets, preprocessing time series, training recurrent neural network models on historical prices, and evaluating the predictions a model produces. Sentiment analysis hooks round out the feature set for traders who want to combine price signals with mood signals. The library has been on GitHub for years and was last updated in 2021, so treat it as a compact, readable reference implementation for learning how deep learning stock prediction pipelines are put together rather than as a maintained production trading system.

Key Features

• Share objects – load OHLCV price history for a ticker from Yahoo Finance in a couple of lines.
• Deep learning models – TensorFlow-based recurrent networks for price prediction and modelling.
• Train/test splitting – built-in evaluation helpers to partition time series data correctly.
• Preprocessing utilities – prepare raw market data for model input without boilerplate.
• Sentiment analysis – hooks for combining market sentiment with price-based signals.
• Documented API – full documentation hosted on Read the Docs with usage examples.

💳 Pricing

Bulbea is free and open source. You install it from source with pip and its dependencies, including TensorFlow, and run everything locally — there is no paid tier, hosted version or account required.

Who Is It For

Python developers, quant hobbyists and students who want a small, hackable codebase for experimenting with deep learning approaches to stock market prediction and for learning how such pipelines are structured end to end.