Xgboost time series prediction method combined with complementation-ensemble empirical mode decomposition

Abstract

The invention discloses an Xgboost time series prediction method combined with complementation-ensemble empirical mode decomposition. The method includes the following specific steps that 1, data is preprocessed; 2, complementation-ensemble empirical mode decomposition of sale-data time series is processed with the complementation-ensemble empirical mode decomposition method; 3, orthogonality factors and a regression model are established through Xgboost; 4, parts outside of the influence of non-orthogonal characteristic factors are fitted through the Xgboost. According to the Xgboost time series prediction method combined with complementation-ensemble empirical mode decomposition, through the orthogonality of intrinsic mode functions, it is guaranteed that a convergence path when the Xgboost searches an optimal model extends from outer space to inner space, then a path is searched from the inner space to be converged to the corresponding intrinsic mode function step by step, outer-space errors and inner-space errors are decreased respectively in the two convergence processes, the orthogonality is also guaranteed, and generalization of the model is finally improved.

Description

technical field [0001] The invention relates to a time series prediction method, in particular to an Xgboost time series prediction method combined with complementary set empirical mode decomposition. Background technique [0002] Time series refers to the sequence of the values ​​of the same statistical index arranged in the order of their occurrence time. In daily production and life, the collected time series changes are often affected by various factors, such as weather conditions, traffic conditions, etc. Conditions, geography, etc. These factors lead to the non-stationary state of the collected time series. Commonly used time series predictive analysis models, such as the autoregressive moving average model, are established on the premise that the time series satisfies the assumption of stationary conditions, so these models are more suitable for analyzing stationary data. In order to better analyze complex non-stationary data, signal analysis can be used to properly ...

AI Technical Summary

Problems solved by technology

Empirical mode decomposition is a signal analysis method, which can decompose an unsteady signal into a series of single modes, that is, signals that satisfy the stationary state. However, in some current prediction methods based on empirical mode decomposition, only After simply decomposing the data, modeling and predicting each component, the predicted data at the next moment is the sum of the predicted values ​​of the components, which does not take into account the existence of "mode" in the intrinsic mode function generated by the empirical mode decomposition method. Stat...

Images