Search

In the stock market, accurate trend judgment and reasonable asset distribution are effective ways to obtain ideal return. However, the real stock market is affected by the objective economic environment, investors' expected return and other potential factors, which makes the classical portfolio strategy face more challenges and pressures. How to build a reliable portfolio strategy in an uncertain environment will be a scientific problem worthy of in-depth discussion. To address this issue, this paper combines machine learning with rough set to establish a new rough set theory prediction model, quantitatively dividing the stock data into three categories and targetedly predicting the future trend according to the complexity. Based on the proposed prediction model, a new portfolio strategy is proposed by integrating the mean-variance model. Firstly, for reducing the volatility and noise of the original data of stock price, outlier processing (OP) and wavelet denoising (WD) are utilized. Secondly, for the sake of pertinently forecasting the future trend of different characteristic stock price, a three-way prediction (TWP) decisions approach is constructed based on multiscale permutation entropy (MPE), probabilistic rough set (PRS), variational modal decomposition (VMD) and deep learning. Finally, 20 stocks of Shanghai and Shenzhen stock exchanges are taken as research samples to verify the scientificity and rationality of the portfolio strategy. The results show that the proposed approach not only provides scientific support and reference for investors' investment decisions, but also provides a new investment strategy theory and method for the investment decisions of the stock market. [ABSTRACT FROM AUTHOR]