MNFS-FPM: A novel memetic neuro-fuzzy system based financial portfolio management

Portfolio management consists of deciding what assets to include in a portfolio given the investor's objectives and changing market and economic conditions. The always difficult selection process includes identifying which assets to purchase, how much, and when. This paper presents a novel memetic neuro-fuzzy system for financial portfolio management (MNFS-FPM) which emulates the thinking process of a rational investor and generates the optimal portfolio from a collection of assets based on a chosen investment style. The system consists mainly of two modules: the generic self-organizing fuzzy neural network realizing Yager inference (GenSoFNN-Yager), to predict the expected return of each asset, and a memetic algorithm using simplex local searches (MA-NM/SMD) to determine the optimal investment weight allocation for all assets in the portfolio. Experimental results on Dow Jones industrial average (DJIA) stocks show that the proposed system yields better performance compared to that of existing financial models: statistical mean-variance analysis and capital asset pricing model (CAPM).