Real-time time series predicting method and device

Abstract

The invention discloses a real-time time series predicting method and device. The real-time time series predicting method comprises the steps that a period is divided into multiple sections, and a model is established for each section with longitudinal time as a dimension; according to historical data, modeling is conducted on differential values of the same moment of different periods through the Gaussian model; during prediction, according to the moment to be predicted, the corresponding model is called, and a reference value of predication data is output, wherein the reference value is obtained by adding the data of the same moment of the last period and the difference value output by the corresponding model together. According to the real-time time series predicting method and device, due to the fact that modeling is conducted on the longitudinal time dimension, the complexity of each model is lowered.

Description

technical field [0001] The invention relates to the technical field of signal processing, in particular to a time series real-time prediction method and device. Background technique [0002] The time series forecasting method is based on the ordered observation data sets associated with the time sequence, and uses the stochastic process theory and mathematical statistics methods to study the statistical laws obeyed by the data sets, so as to speculate on the data development trend and guide the solution of practical problems. Time series forecasting method has been widely used in industry, address, ecology, economy, meteorology, medicine and other fields. Time series forecasting techniques can be roughly divided into: [0003] 1. Traditional linear time series forecasting techniques. In 1968, Box and Jenkins proposed a relatively complete set of time series modeling theory and analysis methods. These classical mathematical methods predict time series by establishing stoch...

AI Technical Summary

Problems solved by technology

[0007] However, most of the existing time series forecasting algorithms have complex models in order to obtain more accurate forecast results, and the training time of the models is too long, which affects the subsequent use of the models

Moreover, most of the existing time series forecasting algorithms require a large amount of training data in order to obtain more accurate forecasting results

In addition, most existing time series forecasting algorithms only consider a single time dimension

Images