A Hybrid Model for Portfolio Optimization Based on Stock Price Forecasting with LSTM Recurrent Neural Network Using Cardinality Constraints and Multi-Criteria Decision Making Methods (Case study of Tehran Stock Exchange)

Due to the dynamic trend of stock prices and the volatile nature of the market, asset price forecasting plays a key role in creating an efficient strategy. In addition, the results of price forecasting are a prerequisite for creating a portfolio with an optimal structure. Accordingly, the purpose of this research is to provide a hybrid model to help investors in selecting optimal portfolios. For this reason, ten top preferable industries have been selected among the active industries of Tehran Stock Exchange using the Improved Analytical Hierarchy Process method, Then, the price of selected active industries' stocks has been predicted daily, monthly, bi-annually, and annually, using a Long Short Term Memory Recurrent Neural Network. In the next step, three portfolios with different time horizons have been selected by using the Combined Compromise Solution method, and finally, optimal weights have been determined and an efficient frontier has been drawn using Mixed-Integer Quadratic Program and Branch and Cut Algorithm based on Limited Asset Markowitz Model. According to the results of this research, the proposed model gives higher returns to investors due to the risk in constituting portfolios with specified time horizons in contrast to traditional approaches.