Efficient mobility management involves micromobilityprinciples. The performance of the Hierarchical MobileIPv6 (HMIPv6) protocol, a representative micro-mobilityapproach, is affected by the Mobility Anchor Point (MAP)selection. In this paper, we propose a new selection method based on a prediction of the future movements of Mobile Nodes (MNs). The proposed algorithms exploit the information about the future availability of MAPs and choose those MAPs that assure a better service. An improvement to the evaluation methodology is also proposed. The algorithms are compared to each other notonly in synthetic but also in realistic internet topologies, which has not been a practice in the past. The simulation results show promising improvements in terms of distance from chosen MAPs and frequency of MAP changes. Moreover, we showed that, for perceivable improvement of MAP selection, absolute accuracy of movement prediction is not required. As pioneers in the mobility management analysis in realistic environment, we ascertain that offering MAP services over more than one Autonomous System (AS) proves beneficial.