Abstract: Predictive range aggregate (PRA) queries are one of the important researching areas in the moving object database. In this paper an efficient prediction technique, PRA-tree, is presented for range aggregation of moving objects. PRA-tree splits the velocity domain regularly, and classifies moving objects into different velocity buckets by their velocities. Then a TPR-tree, which is based on the TPR-tree structure and added with aggregate information in intermediate nodes, is used to index the moving objects in each buckets, thus reducing the disk accesses of PRA queries. A PRA-tree is supplemented by a hash index on leaf nodes, and uses bottom-up delete algorithm, thus having a good update performance and concurrency. Also developed for the PRA tree is an enhanced predictive range aggregate (EPRA) query algorithm which uses a more precise branch and bound searching strategy, reducing the disk I/O greatly. Experimental results and analysis show that the EPRA algorithm for PRA-tree has a good query performance and outperforms the popular TPR\+*tree index.