Online reinforcement learning transaction system and method

Abstract

The invention discloses an online reinforcement learning transaction system and method. The system comprises a collection module used for collecting transaction data of a transaction market, a feature mapping module used for performing continuous space mapping on the transaction data to obtain a reinforcement learning state, a reinforcement learning model used for performing Q function approximation in an iterative mode and updating model parameters, an output module used for outputting a Q value of each operation according to the reinforcement learning model and a reinforcement learning state obtained after each iteration and extracting a maximum value in Q values of all operations, and a calculation module used for evaluating a transaction effect according to the maximum value in the Q values of all the operations. According to the system, complex signals of the transaction market can be learned, and the market change can be learned in real time and adapted, so that the practicality and reliability of the system are improved; and the system is simple and easy to realize.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to an online reinforcement learning transaction system and method. Background technique [0002] With more and more trading data in the market, the need for rapid analysis of large amounts of data is becoming more and more urgent, especially in some high-frequency trading, relying on human beings to issue trading orders is likely to lag behind and lead to loss of profit opportunities. Therefore, computer technology is applied in transactions, so as to rely on the program to automatically analyze the existing data to discover arbitrage opportunities and issue trading instructions to make profits. [0003] In related technologies, the application of programmatic transactions mainly includes the following two aspects: [0004] 1. For the analysis of fundamental information, the program can automatically analyze documents such as the company's financial statements, and obtain valuabl...

AI Technical Summary

Problems solved by technology

However, there are still many deficiencies in the existing programmed trading algorithms. For example, most of them require the designer to have rich trading experience, refine the trading experience, come up with trading ideas, and then use the program to realize it, which is not only limited by the trading personnel. trading experience, and it is not conducive to fully analyze market data to discover more arbitrage opportunities

Images