Forecasting and Selecting the Optimal Investment Portfolio in Socio-economic Systems Using Genetic and Levenberg Neural Networks

The most important objective of any investor is to improve efficiency and reduce the investment risks. Due to the importance of this issue, numerous studies have been done in this field. But in this study, prediction of return and risk covariance has been studied through Genetic-Levenberg neural network, which programmed in matlab software. To present the method validation, the efficiency of top stock exchange companies in Tehran has been evaluated. Results indicated that using this method ensures efficiency with high accuracy and minor errors in the shortest possible time by calculating the mean squared error. An exact formula was obtained which specifies the competency of optimizing a company planning to enter the portfolio. This procedure shows a more efficient performance in comparison to other algorithms and neural networks in terms of high precision, lower error, and less calculation time.

Keywords:

portfolio optimization , return forecast , Levenberg , genetic algorithm , artificial neural network