This paper studies deeply the technology of mobility prediction, analyzes the shortcomings of the existing clustering algorithms, and presents an adaptive clustering algorithm based on orbit prediction in the mobile MANET. According to a node’s moving history, its future moving track can be predicted using the algorithm. Besides, the stability of the connections between adjacent nodes can be estimated, and then the network topology based on the connection stability can be constructed by the algorithm.Finally, all the nodes in MANET are divided into several clusters by the graph coloring technique. The algorithm maintains the stability for the clustering framework and effectively prevents the interruption of communications, thus compensates for the flaws of the previous clustering algorithms.