A Chaos Time Series Prediction Method of Generalized Phase Space

Aiming at the limitation of the regression prediction method for mine gas emission, a chaotic time series prediction method for reconstruction of generalized phase space is proposed in this paper. This method is used to construct phase spaces for time series with unobvious chaotic characteristics and fit equivalent chaotic attractors by RBF and BP network in order for prediction researches by calculating saturated embedding dimensions and designing main neural network parameters. A prediction method for stabilizing the neural network prediction result is brought forward and wavelet denoising is made for original signals to improve the prediction precision. This method can effectively overcome the deficiency of record data types, with stronger practicability, then the simulation experiment on Mackey-Glass time series and actual mine gas emission has proved the feasibility and effectiveness of this method.