Research On Investment Portfolio Management Based On Copula Entropy Stock Selection And Integrated Neural Network Predictio

As one of the intermediary means of financial integration,the capital market plays an important role in China’s economic life.With the advancement of China’s financial reform,more and more enterprises raise funds through listing to solve the capital bottleneck problem of their own development.At the same time,identifying good enterprises and participating in the growth of enterprises through investment means to obtain investment returns has also become the investment demand of more and more people.Since the stock price is the result of the company’s operation and other factors,its trend often presents a complex situation,which brings great challenges to the risk control of stock investment.In order to effectively control investment risks,investors are more inclined to use investment portfolios to participate in investment practices.Two questions that must be solved in constructing a portfolio are:(1)Which stocks are selected as investment targets for portfolio construction?(2)How to realize investment returns based on the constructed portfolio? On the basis of fundamental stock selection,this paper introduces entropy to screen the stock pool from the perspective of system complexity to reduce the investment volatility of the portfolio,so as to achieve the purpose of controlling investment risk.After constructing the portfolio,this paper introduces a deep learning algorithm to predict the portfolio,and improves the winning rate of investment by improving the prediction accuracy,so as to achieve the purpose of improving investment returns.In order to ensure the good performance of investment targets,this article selects stock pools from three perspectives: profitability,growth and valuation reasonableness,and the data covers the time period from January 1,2016 to December 31,2021.Secondly,through the calculation of entropy value,three groups of portfolios with different risk levels are selected for comparative experiments,and the weights under the lowest risk level in the portfolio are determined according to the Monte Carlo simulation results,and the portfolio is established accordingly.After establishing the optimal investment portfolio,this paper introduces the convolutional long short-term memory neural network(CLSTM)to predict the investment portfolio,and then converts the prediction information into timing trading signals,and verifies the practical value of the proposed model by comparing the differences in investment effects between different models.The empirical results show that increasing the entropy value of the portfolio can reduce the degree of correlation between different investment targets in the portfolio,increase the systematic complexity of the portfolio,thereby reducing the investment volatility in the portfolio,and is conducive to the control of investment risks.Referencing convolutional long short-term memory neural networks can improve the forecast accuracy of portfolios,thereby improving investment returns.The optimal portfolio risk level screened in this paper can be reduced from 0.2925 to 0.2654;the mean squared error of the forecast model proposed in this paper is 1.4%,the average absolute value error is 9.1%,and the average Jedi percentage error is 2.196,and the prediction accuracy is better than that of the comparison model;the accuracy of the trading signals in this paper is 72.6%,the accuracy rate is 75.58%,the annualized return of the test set is 197.34%,and the annualized volatility is 144.2%.The annualized Sharpe ratio is 1.1017.From the empirical results,the model proposed in this paper can help investors reduce investment risks,improve investment returns,and have good application value.