Portfolio Construction Using Neural Networks and Multiobjective Optimization

In recent times, financial markets have been increasingly affected by significant volatility and uncertainty. Given this backdrop, it is beneficial for investors to explore a broader range of asset classes when constructing their financial portfolios. This paper examines the concept of a mixed portfolio from 10 different assets. A technical analysis on the selected data has been conducted using Excel and MATLAB. Subsequent price movements of the chosen instruments were then predicted for the subsequent period employing NARXNN method. This led to the evaluation of the expected rates of return for these financial instruments. The estimations were then blended into an optimal risk portfolio, which maximizes return and minimizes risk, based on the solution to a multi-objective optimization problem. To assess risk, the standard deviations of the rates of return and the correlation matrix between the return rates of the considered financial instruments were utilized.