In this paper, a target prediction scheme for handoff in NGI is presented based on both statistical and instant information. At first, it adopts a statistical prediction scheme to predict handoff target, using fuzzy decision tree which is established based on the prior handoff historical records. If there exist several predicted handoff targets, an instant prediction scheme is initiated to compute the probability for each predicted handoff target being visited based on the user's current position, movement rate, direction and the average call duration time, choosing the one with the maximum visited probability as the handoff target. Simulation results have shown that the proposed scheme has higher prediction hitting rates under different user movement rates compared with those schemes simply based on statistical or instant information.