Abstract: This paper introduced ARIMA prediction mechanism in weighted clustering algorithm (WCA). During routing maintaining process, the ARIMA is used to predict the network node location. Using the established ARIMA model, the algorithm is able to predict geographical position of the node at next time. In this way, it can calculate aggregate holding time of the nodes. Then, the predicted aggregate holding time is compared with time warning threshold, if clusto structure will be unstable, route recovery process will be activated before the link fails. And it will search new routing in order to avoid frequent link failures. Thus, the influence to the routing protocols brought by the dynamic changes of network topology can be reduced, and the stability of the cluster structure will be maintained. The simulation results show that, compared with LOWID and RLWCA not joined the forecasting mechanism, the proposed ARP-LWCA algorithm can dramatically improve the network packet delivery rate,reduce the network normalized overhead and the number of routing interruptions significantly. So, the network performance is improved.

Key words: Ad-hoc networks, ARIMA prediction, Cluster algorithm, CBRP