Dinamik beklenti-değişinti problemi için üç amaçlı bir reformülasyon

The classical mean-variance problem aims to find a portfolio that minimizes a linear combination of the expectation and the variance of the terminal wealth. The dynamic version of the problem is known to be time-inconsistent in the classical sense, which makes the scalar dynamic programming approach inapplicable. By decomposing variance into two separate objectives, we introduce a tri-objective formulation in a discrete-time framework that generalizes the scalar problem and can reduce to the original setting. Using a less restrictive concept of time-consistency in a vector-valued sense, we show that the new formulation is time-consistent. Following the literature on set optimization, we develop a set-valued dynamic programming principle with the upper image of the vector-valued problem used as a value function. Finally, we reduce the generalized solutions of the formulation to the classical mean-variance problem using the minimal points of the three-dimensional upper images. We compute portfolios that are optimal for the initial mean-variance problem, and that remain time-consistent with respect to the tri-objective formulation.