Evaluation and Analysis of an LSTM and GRU Based Stock Investment Strategy

Confronted with an extremely complicated and volatile external environment, it is such a tremendous challenge for researchers and investors to predict the stock market prices. To address the challenge, this paper proposes three steps for stock investment. The stock selection is based on a special ratio which is forward PE divided by trailing PE. This ratio can better evaluate the growth of individual stocks. The research found that stocks that have low PE ratios show strong growth in the price prediction part. Two deep learning-based stock market prediction models are proposed to predict the tendency. LSTM and GRU models are separately adopted to predict future trends of stock prices based on the price history. The experimental results show that the GRU model can improve prediction accuracy and reduce time delay, compared to the consequences of the LSTM model. After determining the scope of investment, to reduce the risk of investment in the stock market, get a higher or more stable rate of return, and achieve a good investment, this study calculated the correlation between these stocks’ changes and then optimize the asset allocation. Monte Carlo model and SLSQP model are used to get the correlation between stocks and both of them to give the respective optimal portfolio. From the latter’s results, the diversity of portfolios decreases with the optimization of asset allocation.