wangshub/RL-Stock
RL Stock
RL-Stock is a system for executing deep reinforcement learning trading bots designed to automate stock trading and optimize financial profit strategies. It includes a financial market data pipeline for fetching and normalizing historical price and volume data and a policy gradient trading optimizer to refine continuous action outputs.
The project provides a simulation environment to measure the performance of trading strategies against initial capital using unseen test data. It employs policy gradient reinforcement learning algorithms to optimize trading actions and improve overall portfolio performance.
The system covers the full workflow of algorithmic trading, including data acquisition from public APIs, feature normalization, and the mapping of portfolio value changes into reward signals. It further supports the evaluation of trading logic on historical data to measure profitability and risk.
Features
- Automated Stock Trading - Uses reinforcement learning to automate stock trading decisions to maximize financial returns.
- Policy Gradient Optimizers - Uses policy gradient reinforcement learning algorithms to refine continuous action outputs in stock trading.
- Portfolio Optimization Algorithms - Applies policy gradient algorithms to optimize trading actions and improve overall portfolio performance.
- Market-Based Reward Signals - Maps portfolio value changes into scalar reward signals to quantify trading success for the learning agent.
- Reinforcement Learning Trading Frameworks - Implements a deep reinforcement learning system to automate stock trading and optimize profit strategies.
- Financial Market Simulations - Models the stock market as a Markov Decision Process to simulate agent interactions with historical price data.
- Algorithmic Trading Simulators - Provides a simulation environment to measure trading strategy performance against initial capital using unseen test data.
- Trading Strategy Backtesting - Simulates trading logic on historical data to measure profitability and risk before live deployment.
- Feature Scale Normalization - Transforms raw market price and volume data into a fixed numerical range to ensure stable model convergence.
- Market Data Transformation Pipelines - Provides a pipeline for fetching and transforming raw market data into normalized inputs for predictive models.
- Financial Data Processing - Collects and normalizes historical price and volume data for machine learning model training.
- Market Data Normalizers - Converts raw financial metrics into a consistent numerical range to prepare data for model training.
- Financial Data Retrieval APIs - Provides utilities to retrieve historical price, volume, and turnover data from public financial APIs.
