Portfolio selection for neural network using energy networks in Tehran Stock Exchange

Optimization problems are one of the most interesting, important, and popular fields of financial mathematics. A better portfolio optimization model can help investors to earn more sustainable profits. The existing literature shows that the performance of traditional mean-variance portfolio strategies is not suitable. To address this issue, this study uses a multilayer perceptron neural network and a convolutional neural network to predict the future direction of stock prices.

We compare the prediction accuracy of these two methods and enter the outputs of each method with higher accuracy into the proposed model. Then, given the future direction of stock prices, we propose an efficient stock selection scheme for investors. We also test the proposed stock selection scheme and investment strategies using the Tehran Stock Exchange index components as test cases.

The experimental results show that the proposed stock selection scheme can effectively improve the performance of all investment strategies. In addition, the proposed investment strategy outperforms the traditional minimum global variance investment strategy.

Originality/Value: 

This research provides an innovative framework for portfolio selection based on deep learning networks. These networks are key to increasing investment efficiency, risk management, and decision-making in the Iranian capital market and provide an advanced model for similar markets.