A Review of Time Series Prediction Methods Based on Deep Learning
Bin Tian, Zhou Sun, Hanyu Hong · 2023
Time series is a way of presenting statistical indicators in a particular order based on the observation of a specific process at a certain sampling frequency. This collection of values is reliant on the two adjacent values and is used for predictive analysis which estimates and predicts the future development of an object based on known information. Time series forecasting aims to analyze the development trend of an object by sequencing the information already available according to a certain time, so as to make reasonable speculations about the possible state of the future period. Sequence prediction is widely used in areas such as finance, traffic flow, electricity, and operational load of various systems. Deep learning is a technique that utilizes the original sequence and other relevant input information without feature engineering to learn the implicit expression of the time series through various model structures. Further research directions on time series prediction methods based on deep learning are highlighted in this paper.