基于机器学习的投资组合研究

Research on Investment Portfolios Based on Machine Learning

股票市场的随机性使得如何利用已知信息构造投资组合并在规避风险的前提下获得最大收益成为人们关注的重要问题。使用熵权-TOPSIS法对指标进行赋权并据此对沪深300成分股进行排序筛选。基于随机森林模型和LSTM(longshort-termmemory)模型对股票收盘价进行预测,选择月收益率排名前7的股票利用均值-方差模型进行投资组合构建,分析不同预测模型下投资策略的收益情况,最终通过实验验证选择LSTM预测模型得到的投资组合能够获得更高的收益,且具有较高的回报稳定性。

The randomness of the stock market makes how to use known information to construct an investment portfolio and achieve maximum returns while avoiding risks an important issue of concern.The entropy weight-TOPSIS method was used to assign weights to the indicators and rank and select the CSI 300 constituent stocks.The random forest model and LSTM model were used to predict stock closing prices.The top 7 stocks in monthly returns were selected for portfolio construction using the mean-variance model.The returns of investment strategies under different prediction models were analyzed.The experiments ultimately verify that the investment portfolio selected using the LSTM prediction model achieves higher returns and has greater return stability.