基于邻近相点聚类分析的多变量局域多步预测

Multivariate local multi-step prediction based on cluster analysis of adjacent phase points

针对高维混沌复杂系统的多步预测问题,提出了一种基于邻近相点聚类分析的多变量局域多步预测模型。首先对于多变量邻近相点的选取,结合邻近相点多步回溯后的演化规律和变量间的关联信息对演化轨迹的影响,提出了一种新的多变量演化轨迹相似度综合判据;然后针对选取全局最优邻近相点耗时长的缺点,提出了一种基于邻近相点聚类分析的新方案来降低多步预测时间,提高预测效率。最后通过Lorenz混沌数据仿真实验,表明该模型具有优良的预测性能。

In order to solve the problem of multi-step prediction for high dimensional chaotic complex systems, this paper proposed a multivariate local multi-step prediction model based on cluster analysis of adjacent phase points. First, it considered the evolution rule of adjacent phase points after multi-step backtracking and the influence of related information on evolutionary trajectory for selecting adjacent phase points. So this paper proposed a new multivariate evolutionary trajectory similarity criterion. Then, the traditional model took a long time to select the optimal phase points. To overcome this shortcoming, it proposed a new method based on cluster analysis of adjacent phase points. This method can reduce the multi-step prediction time and improve prediction efficiency. Finally, the simulation results of Lorenz chaotic data show that this model has good prediction performance.