In this paper, we study the portfolio optimization model with transaction costs under the assumption that there exist admissible errors on expected returns and risks of assets. We propose an admissible efficient portfolio selection problem and design a neural network for the proposed problem. The presented neural network framework guarantees to obtain the optimal solution of the admissible portfolio selection problem. The existence and convergence of the trajectories of the network are studied. The Lyapunov stability and globally convergence of the considered neural network are also shown. We provide a numerical example to illustrate the proposed effective approach.