A multivariable time sequence prediction method based on adaptive noise reduction and integrated LSTM

Abstract

The invention discloses a multivariable time sequence prediction method based on adaptive noise reduction and integrated LSTM (Long Short Term Memory). The multivariable time sequence prediction method is used for solving the problems of unstable performance and low prediction precision when a multivariable time sequence with non-stationary, non-linear and noisy characteristics is predicted by anexisting method. The method comprises the following steps: decomposing a noise-containing chaotic multivariable time sequence by adopting a complete set empirical mode decomposition method with adaptive noise to obtain a series of intrinsic mode functions with frequencies from high to low; Distinguishing a noise-containing high-frequency intrinsic mode function from a low-frequency noise-free intrinsic mode function by adopting a permutation entropy thought; Constructing a self-adaptive threshold value and a self-adaptive threshold value function to reduce noise of the noisy intrinsic mode function; Constructing a stacked automatic encoder to extract characteristics of the multivariable time sequence after noise reduction; Constructing a multivariable time sequence weak predictor based onthe LSTM neural network; And constructing an integrated algorithm considering the prediction error of the verification set, and combining a plurality of LSTM weak predictors to obtain a strong predictor.

Description

technical field [0001] The invention belongs to the technical field of computer applications, and in particular relates to a multivariate time series prediction method based on adaptive noise reduction and integrated LSTM. Background technique [0002] In actual production and scientific research, the observed values ​​arranged in chronological order by observing and measuring a certain index or a group of indicators are called time series data. The time series model can fit and learn the time series data. Such as random changes, periodic changes or trend changes. The multivariate time series prediction model is aimed at multiple variable time series, fully considers the relationship between each variable time series, and predicts one or more target time series. Multivariate time series forecasting has been widely used in many fields, such as financial market forecasting, energy forecasting and environmental pollution forecasting, etc. It is of great significance to predic...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a multivariate time series prediction method based on adaptive noise reduction and integrated LSTM, in order to solve the problem of poor performance of existing methods for multivariate time series prediction with non-stationary, nonlinear and noisy characteristics. The problem of stability and low prediction accuracy

Images