Stock ordering method and device based on reinforcement learning

Abstract

The invention provides a stock ordering method and device based on reinforcement learning. The method comprises the following steps: constructing a plurality of state parameters in a stock transactionprocess based on historical snapshot data, wherein the state parameters comprise public state parameters of the whole market and private state parameters of transaction users; constructing a plurality of behavior parameters in the stock transaction process according to a preset corresponding relationship and the state parameters, wherein the behavior parameters comprise stock buying actions and / or stock selling actions at different prices; obtaining a corresponding transaction parameter after each behavior parameter is adopted under each state parameter; generating and training a quality matrix according to the state parameters, the behavior parameters and the transaction parameters; obtaining a current state parameter of a target individual strand, and determining a current behavior parameter with the maximum excitation parameter corresponding to the current state parameter according to the quality matrix; and limiting the price of the target individual share based on the current behavior parameter.

Description

technical field [0001] The present invention relates to the technical field of stock trading, in particular to an order placing method, device, computer equipment and storage medium based on reinforcement learning. Background technique [0002] In the existing stock trading process, when receiving an average price order from a user to buy a certain amount of a certain stock within a long time interval, the TWAP strategy, VWAP strategy or machine learning strategy is usually used to automatically The order is placed within this time interval. The TWAP strategy distributes active transaction orders evenly over time, and the price of the order directly covers the five-level price of the opponent's market in order to be able to complete the transaction. The way of buying the opponent's order completely across the middle price difference will greatly increase the transaction cost. According to the different prices of different stocks, this operation will increase the transaction...

AI Technical Summary

Problems solved by technology

This operation is also a market order operation, so there is still an increase in transaction costs

Furthermore, because the machine learning strategy buys and sells at the end of the range, it is easy to be captured by other participants who can obtain market transaction data, and then form a reverse strategy, further causing adverse effects

Images