Time sequence prediction method based on GRU neural network

Abstract

The invention belongs to the technical field of network information prediction, and discloses a time sequence prediction method based on a GRU neural network. The method comprises the following steps:collecting original data needing to be predicted; carrying out data preprocessing on the collected original data; performing standardization processing on the preprocessed data; performing dimensionraising processing on the original time series data form by using the code; training the input data by using a GRU neural network to obtain a trained time sequence prediction model, and storing the trained time sequence prediction model; predicting the time series data by using a GRU-SES model to obtain a preliminary prediction value; performing secondary exponential smoothing processing on the obtained preliminary prediction data to obtain a final prediction data value; output of prediction results. The prediction method provided by the invention improves the precision of time sequence prediction, and is of great significance to time sequence analysis in industrial production or actual life.

Description

technical field [0001] The invention belongs to the technical field of network information prediction, and in particular relates to a time series prediction method based on a GRU neural network. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: time series are widely present in daily life and production, and the development of everything in the world develops with time, such as in weather, finance, transportation, industry, agriculture, etc. There are variable series formed according to the time dimension in various fields. Changes in daily temperature in weather, changes in passenger traffic in traffic, changes in stock prices in finance, etc., all exist in the form of time series. At the same time, some data in life can also be converted into a time series representation, such as genetic data, etc. Time series are a special form of stochastic process. Time series is a series of numbers that arranges the sta...

AI Technical Summary

Problems solved by technology

[0011] 3) The ARMA model is evolved from the autoregressive model (AR) and the moving average model (MA). The prediction accuracy is improved by mixing AR and MA, but the parameter estimation is more complicated

[0020] (1) The function form used in the existing forecasting models using traditional probability statistics is relatively fixed, and it cannot obtain good results for the high-complexity and nonlinear time series forecasting in practice;

[0021] (2) The existing artificial neural network prediction model has some defects such as parameter optimization and local optimum

For example, in the inventory of commodities, how to determine the inventory capacity of the future warehouse is of great research value. A small inventory capacity is not conducive to the sale of commodities; a large inventory capacity is not conducive to inventory management and increases inventory management costs.

Images