What is this?
FinRL is the original open-source framework for financial reinforcement learning, letting you train deep RL agents (A2C, DDPG, PPO, TD3, SAC) to trade stocks and other assets. It's built around a train-test-trade pipeline split into market environments, DRL agents, and financial applications.
FinRL is the original open-source framework for financial reinforcement learning, letting you train deep RL agents (A2C, DDPG, PPO, TD3, SAC) to trade stocks and other assets. It's built around a train-test-trade pipeline split into market environments, DRL agents, and financial applications.Is this a ready-to-run trading bot, or a framework I build on?It's a framework: you assemble market environments, DRL agents, and financial applications yourself through train.py, test.py and trade.py style scripts, rather than getting a plug-and-play bot.
It's a framework: you assemble market environments, DRL agents, and financial applications yourself through train.py, test.py and trade.py style scripts, rather than getting a plug-and-play bot.How do I get it running?Clone the repository, create and activate a virtual environment, install FinRL, then run the provided scripts to download/preprocess stock data, train the five DRL agents, and backtest them.
Clone the repository, create and activate a virtual environment, install FinRL, then run the provided scripts to download/preprocess stock data, train the five DRL agents, and backtest them.Do I need broker API keys for data, or does it work out of the box?The default tutorial pulls DOW 30 data from Yahoo Finance with no account needed, but most other supported sources like Alpaca, Tushare, Binance, or WRDS are account-specific and require their own API keys.
The default tutorial pulls DOW 30 data from Yahoo Finance with no account needed, but most other supported sources like Alpaca, Tushare, Binance, or WRDS are account-specific and require their own API keys.Should I use this or FinRL-X?FinRL-X is the actively developed, production-oriented successor with live multi-account trading and risk controls; this classic FinRL repo is frozen as the original educational and research framework, so don't expect new production features here.
FinRL-X is the actively developed, production-oriented successor with live multi-account trading and risk controls; this classic FinRL repo is frozen as the original educational and research framework, so don't expect new production features here.Does it run on Windows as well as Mac and Linux?Yes, the README provides install instructions for macOS, Ubuntu, and Windows 10.
Yes, the README provides install instructions for macOS, Ubuntu, and Windows 10.Which RL algorithms come with it?It trains five DRL agents out of the box using Stable Baselines 3: A2C, DDPG, PPO, TD3, and SAC.
It trains five DRL agents out of the box using Stable Baselines 3: A2C, DDPG, PPO, TD3, and SAC.What's the difference between FinRL, FinRL-Meta, and FinRL-X?FinRL-Meta supplies the gym-style market environments and benchmarks, FinRL (this repo) is the classic end-to-end train-test-trade framework for learning and research, and FinRL-X is the next-gen, modular, production-oriented stack for institutions.
FinRL-Meta supplies the gym-style market environments and benchmarks, FinRL (this repo) is the classic end-to-end train-test-trade framework for learning and research, and FinRL-X is the next-gen, modular, production-oriented stack for institutions.How is backtesting evaluated?The backtest script runs the trained agents on held-out trading data and compares their returns against Mean Variance Optimization and the DJIA index, saving a result plot.
The backtest script runs the trained agents on held-out trading data and compares their returns against Mean Variance Optimization and the DJIA index, saving a result plot.What markets can it trade beyond US stocks?It supports 14+ data sources covering US, CN, and Taiwan securities as well as cryptocurrencies through Binance and CCXT, each delivering OHLCV and preprocessed price/indicator data.
It supports 14+ data sources covering US, CN, and Taiwan securities as well as cryptocurrencies through Binance and CCXT, each delivering OHLCV and preprocessed price/indicator data.What are similar projects?qlib, FinGPT, abu, Qbot, machine-learning-for-trading, anti-gambling-trader-tw, awesome-quant, TradingAgents
qlib, FinGPT, abu, Qbot, machine-learning-for-trading, anti-gambling-trader-tw, awesome-quant, TradingAgentsWhich topics is it in?tensorflow2, deep-reinforcement-learning, algorithmic-trading, stock-market, quantitative-finance, fintech, finance, reinforcement-learning, jupyter notebook, pytorch, benchmark, framework
tensorflow2, deep-reinforcement-learning, algorithmic-trading, stock-market, quantitative-finance, fintech, finance, reinforcement-learning, jupyter notebook, pytorch, benchmark, framework