Mean-variance models for portfolio selection with fuzzy random returns (Q1031991)

In this study, portfolio selection problems are formulated and solved in the framework of fuzzy random variables. The underlying bi-objective programming problem in which the maximum return and the minimum variance are considered comes in the form \[ \begin{gathered} \min V\Biggl[\sum^n_{i=1} \xi_i x_i\Biggr],\\ \max E\Biggl[\sum^n_{i=1} \xi_i x_i\Biggr],\\ \text{subject to }x_1+ x_2+\cdots+ x_n= 1,\;x_i\geq 0,\quad i= 1,2,\dots, n,\end{gathered}\tag{1} \] where \(x_i\) denote the investment proportions in securities, \(\xi_i\) are fuzzy random returns of securities and \(E[\cdot]\) and \(V[\cdot]\) are the expected value and the variance of these variables. Fuzzy random variables are treated as mappings defined as \(\Omega\to F\) satisfying some measurability conditions, where \(F\) is a collection of fuzzy variables defined in the possibility space. As (1) includes two conflicting objectives, it is studied a compromise model in which the individual objective functions are weighted with weight coefficients quantifying risk aversion. Here the solution to this problem is determined by running genetic optimization and the use of triangular fuzzy random variables. Two detailed numeric examples are provided.