Nanyang Polytechnic releases quantitative trading master TradeMaster, covering 15 reinforcement learning algorithms

Recently, the quantitative platform family has welcomed a new member, an open source platform based on reinforcement learning: TradeMaster—Trading Master.

#TradeMaster Developed by Nanyang Technological University is a unified, end-to-end, user-friendly quantitative trading platform covering four major financial markets, six major trading scenarios, 15 reinforcement learning algorithms and a series of visual evaluation tools!

Platform address: https://github.com/TradeMaster-NTU/TradeMaster

Background introduction

In recent years, artificial intelligence technology is occupying an increasingly important position in quantitative trading strategies. Due to its outstanding decision-making ability in complex environments, there is huge potential in applying reinforcement learning technology to tasks in quantitative trading. However, the low signal-to-noise ratio of the financial market and the unstable training of reinforcement learning algorithms make reinforcement learning algorithms currently unable to be deployed on a large scale in real financial markets. The specific challenges are as follows:

Development The process is complex and involves a huge amount of engineering, making it difficult to realize
1. The performance of the algorithm is highly dependent on the market state at the time of testing, the risk is high, and it is difficult to systematically evaluate
2. The design, optimization, and maintenance of the algorithm There is a high technical threshold and it is difficult to deploy on a large scale.

The release of TradeMaster provides this field with a software tool, an industry benchmark and an industrial-grade product interface to solve the three challenges above.

TradeMaster’s potential contribution to the deep integration of industry, academia, research and application

TradeMaster Framework

TradeMaster consists ofsix ​​core modules, including the complete process of design, implementation, testing and deployment of reinforcement learning algorithms for quantitative trading , below we will introduce it to you in detail:

# The framework structure of the TradeMaster platform

Data module: TradeMaster provides long-term multi-modal (K-line and order flow) financial data at different granularities (minute level to daily level), covering four major markets: China, US stocks and foreign exchange.

Preprocessing module: TradeMaster provides a standardized financial time series data preprocessing pipeline, including 6 steps: 1. Data cleaning 2. Data filling 3. Regularization 4. Automatic features Discovery 5. Feature embedding 6. Feature selection

Simulator module: TradeMaster provides a series of data-driven high-quality financial market simulators, supporting 6 mainstream quantitative trading tasks: 1 . Currency trading 2. Portfolio management 3. Intraday trading 4. Order execution 5. High-frequency trading 6. Market making

Algorithm module: TradeMaster implements 7 latest reinforcement learning-based trading algorithms (DeepScalper, OPD, DeepTrader, SARL, ETTO, Investor-Imitator, EIIE) and 8 classic reinforcement algorithms (PPO, A2C, Rainbow, SAC, DDPG, DQN, PG, TD3). At the same time, TradeMaster introduces automated machine learning technology to help users efficiently adjust the hyperparameters of training reinforcement learning algorithms.

Evaluation module: TradeMaster implements 17 evaluation indicators and visualization tools from 6 dimensions: profitability, risk control, diversity, interpretability, robustness, and universality Systematic evaluation. The following are two examples:

Radar chart indicating profitability, risk control, and strategy diversity

Financial time series data visualization

Running process pseudocode

TradeMaster is based on object-oriented programming ideas, encapsulates different functional modules, realizes functional decoupling and encapsulation of different modules, and has good scalability and reusability. The specific process includes the following 6 steps:

Test result

Based on the Dow Jones 30 Index Taking the classic task of investment portfolio as an example, the EIIE algorithm achieved stable positive returns and a high Sharpe ratio on the test set:

##TradeMaster Tutorial

TradeMaster provides a series of different Reinforcement learning algorithm tutorial for trading tasks, presented in the form of Jupyter Notebook to facilitate users to get started quickly:

For details, see: https://github.com/TradeMaster-NTU/TradeMaster/tree/1.0.0/tutorial

The above is the detailed content of Nanyang Polytechnic releases quantitative trading master TradeMaster, covering 15 reinforcement learning algorithms. For more information, please follow other related articles on the PHP Chinese website!