Investment portfolio management method based on deep reinforcement learning

Abstract

The invention discloses an investment portfolio management method based on deep reinforcement learning. The method comprises the following steps: constructing a neural network model of a cryptocurrency transaction scene based on convolution, deep convolution, an extrusion and excitation module, a residual block and a gate circulation unit; training a neural network model based on convolution, deep convolution, an extrusion and excitation module, a residual block and a gate circulation unit to optimize the parameters of the neural network model; loading trained neural network model parameters based on convolution, deep convolution, an extrusion and excitation module, a residual block and a gate cycle unit, and receiving historical price data of cryptocurrency; and obtaining the asset allocation weight at the beginning of the next transaction cycle through the neural network model based on the convolution, deep convolution, extrusion and excitation module, the residual block and the gate circulation unit, and adjusting the allocation of assets in the cryptocurrency market according to the asset allocation weight, thereby obtaining an optimal investment strategy.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to an investment portfolio management method based on deep reinforcement learning in machine learning. Background technique [0002] With the development of artificial intelligence technology, reinforcement learning algorithms have been applied to the financial field. At present, by building a suitable interactive environment, the reinforcement learning model based on neural network has been preliminarily applied to asset management, such as the method shown in the application number 201910321426.X. [0003] However, because the action space in asset management is too large, a single convolution cannot fully explore it, so the benefits of a single convolution model are not very good. In the convolution model of asset management, action is defined as the asset allocation weight determined at the beginning of each transaction cycle. Under this definition, a discrete a...

AI Technical Summary

Problems solved by technology

This seriously affects the application value and generalization ability of the model

Images