Depth decoupling time sequence prediction method

Abstract

The invention discloses a depth decoupling time sequence prediction method, which comprises the following steps of: 1) preprocessing given time sequence data, and constructing a training data set; 2) capturing a global change mode shared by a plurality of time sequences by using a vector quantization global feature encoder; 3) utilizing a local feature encoder to capture a specific local change mode of each of time sequences, wherein each time sequence has a group of specific local feature encoder parameters which are generated by a self-adaptive parameter generation module; and 4) inputting the outputs of the global and local feature encoders into a decoder for prediction. According to the method, the dynamic nature of the time sequences is decoupled into the global change mode and the local change mode, modeling is performed respectively, the problems that an existing model cannot fully utilize shared knowledge in a data set and cannot fully model a specific local change mode of a single time sequence can be solved, the prediction precision is improved, and the method has wide application prospects in the fields of traffic prediction, supply chain management, financial investment and the like.

Description

technical field [0001] The invention relates to the field of time series data prediction, in particular to a deep decoupling time series prediction method. Background technique [0002] Time series widely exist in fields such as transportation, electric power, medical treatment and finance. Time series forecasting (that is, predicting the observed value at a certain moment in the future based on the observed value of a period of history) is an important research topic in data mining. In today's big data era, a single time series often does not exist in isolation, and the data set usually contains multiple correlated time series, which have global (shared by multiple time series) and local (specific to a single time series) changes model. Such as figure 1 As shown, the road usage time series of all roads have the same period (24 hours), and have morning peak and evening peak, that is, the global change pattern; road 1 has a slight morning and evening peak, road 2 has an ob...

AI Technical Summary

Problems solved by technology

However, matrix factorization acts on the feature space and cannot capture complex global variation patterns

Images