Predictive modeling of large-scale sequential curves based on clustering

Traditional approach to predict large-scale sequential curves is to build model separately according to every curve, which causes heavy and complicated modeling workload inevitably. A new method is proposed in this paper to solve this problem. By reducing model types of curves, clustering curves and modeling by clusters, the new method simplifies modeling work to a large extent and reserves original information as possible in the meantime. This paper specifies the theoryand algorithm, and applies it to predict GDP curves of multi-region, which confirms practicability and validity of the presented approach.