Chaotic time series prediction based on phase space reconstruction and LSSVR model

摘要

At present chaotic time series prediction methods are mainly based on reconstructed phase space. However, when the space is reconstructed, two parameters must be determined in advance, they are embedding dimension and delay time. To this problem in the paper authors first introduces the minimum differential Entropy ratio principle to determine the embedding dimension and delay time, and advantage of this method is two parameters simultaneously is solved. Secondly, the phase space can be reconstructed by using the known embedding dimension and delay time. Chaotic time series can be predicted using well-established LSSVR model in the reconstructed phase space. Finally, in MATLAB2009b environment, the algorithm is verified through the Mackey-Glass time-series data and the actual gas emission time-series data. The results show that the geometric meaning is clear and program is simple by minimum differential Entropy ratio principle to determine the embedding dimension and delay time. High time-series prediction accuracy is obtained in this reconstructed phase space, and the same high accuracy also can be obtained in short-term prediction the mining face gas emission.