The decentralized adaptive prescribed performance finite-time tracking control problem is studied for a class of nonlinear interconnected large-scale systems. A new prescribed performance control design method is proposed combining the neural network adaptive technology, the practical finite-time control theory and the prescribed performance control method. The problem of that the traditional prescribed performance method is hard to realize decentralized control is solved. The designed controllers can guarantee that the tracking errors of all the subsystems in the large-scale system are constrained by a finite-time performance function, and the tracking errors can converge to a predetermined neighbourhood of the equilibrium within an arbitrarily given settling time. And all of the signals in the closed-loop large-scale system are practically finite-time stable. Especially, the settling time is irrelevant to the initial system states. Two simulation examples are given to illustrate the effectiveness and superiority of the proposed control method.