Adaptive neuro fuzzy inference system based on portfolio optimization problem

The main purpose of portfolio optimization determining the securities components, stocks, mutual funds, etc. that will best meet the expectations of investors, is to increase return while reducing risk. Since the objective function of the standard portfolio optimization problem has a quadratic form, the solution can be reached with quadratic optimization techniques. The problem turns into a mixed integer optimization problem with the inclusion of the cardinality constraint. To solve this type of problems, hybridized methods can be preferred. The aim of this study is to solve the cardinality constrained portfolio optimization problem with a network based on fuzzy inference system. Firstly, portfolio optimization algorithm based on nonlinear neural network proposed for standard portfolio optimization using neural networks. Secondly, portfolio optimization based on fuzzy inference system, which is a hybrid method for cardinality constrained portfolio selection method, where genetic algorithm and nonlinear neural networks are used together and the coefficient of variation is inverse proposed algorithms for cardinality constrained portfolio optimization. Methods based on this data developed were evaluated for BIST-30 data set. BIST-National all stock market data, analyses were made according to sectors. When the results are compared, it was observed that the algorithms proposed based on hybrid methods are preferable for portfolio optimization.