Abstarct: Optimization problems comprise one of the interesting and popular fields in financial mathematics. The optimal portfolio selection, asset allocation, risk management and option pricing models as more efficient and more accurate models has defined different linear and nonlinear optimization problems. In recent decades, optimization methods that have been developed baxsed on the artificial intelligence, have achieved a great success in effective and efficient solving of optimization problems. Methods such as genetic algorithms, tabu search, simulated annealing, neural networks and etc, well illustrated their ability to solve major scientific problems. Available special rates on neural network models has provided their use in a wide range of research. Among them, we could pointed to the ability to learn and improve performance baxsed on input data, Ability to perform parallel calculations and independently from the initial point of this model, unlike the classic models. In this thesis, using a efficient neural network model we will discuss solving of the fuzzy portfolio selection where every guaranteed interval belongs to a certain class of fuzzy variables. Using a suitable Lyapunov function, we will prove that the expressed neural network model is asymptotically stable and globally convergent to the optimal solution of the original problem. Finally, the results and recommendations for future work will be presented.