Time-Series Prediction Based on Double Pyramid Bidirectional Feature Fusion Mechanism

Na WANG; Xianglian ZHAO

doi:10.1587/transfun.2022EAP1081

Time-Series Prediction Based on Double Pyramid Bidirectional Feature Fusion Mechanism

Na WANG, Xianglian ZHAO

Full Text Views

5

Cite this

Summary :

The application of time-series prediction is very extensive, and it is an important problem across many fields, such as stock prediction, sales prediction, and loan prediction and so on, which play a great value in production and life. It requires that the model can effectively capture the long-term feature dependence between the output and input. Recent studies show that Transformer can improve the prediction ability of time-series. However, Transformer has some problems that make it unable to be directly applied to time-series prediction, such as: (1) Local agnosticism: Self-attention in Transformer is not sensitive to short-term feature dependence, which leads to model anomalies in time-series; (2) Memory bottleneck: The spatial complexity of regular transformation increases twice with the sequence length, making direct modeling of long time-series infeasible. In order to solve these problems, this paper designs an efficient model for long time-series prediction. It is a double pyramid bidirectional feature fusion mechanism network with parallel Temporal Convolution Network (TCN) and FastFormer. This network structure can combine the time series fine-grained information captured by the Temporal Convolution Network with the global interactive information captured by FastFormer, it can well handle the time series prediction problem.

Publication: IEICE TRANSACTIONS on Fundamentals Vol.E106-A No.6 pp.886-895

Publication Date: 2023/06/01

Publicized: 2022/12/20

Online ISSN: 1745-1337

DOI: 10.1587/transfun.2022EAP1081

Type of Manuscript: PAPER

Category: Digital Signal Processing

Authors

Na WANG
Nanjing University of Aeronautics and Astronautics,Nanjing Audit University Jinshen College
Xianglian ZHAO
Nanjing University of Aeronautics and Astronautics

Keyword

time-series prediction, FastFormer, temporal convoluti on network, double pyramid, feature fusion mechanism

Cite this

Copy

Na WANG, Xianglian ZHAO, "Time-Series Prediction Based on Double Pyramid Bidirectional Feature Fusion Mechanism" in IEICE TRANSACTIONS on Fundamentals, vol. E106-A, no. 6, pp. 886-895, June 2023, doi: 10.1587/transfun.2022EAP1081.
Abstract: The application of time-series prediction is very extensive, and it is an important problem across many fields, such as stock prediction, sales prediction, and loan prediction and so on, which play a great value in production and life. It requires that the model can effectively capture the long-term feature dependence between the output and input. Recent studies show that Transformer can improve the prediction ability of time-series. However, Transformer has some problems that make it unable to be directly applied to time-series prediction, such as: (1) Local agnosticism: Self-attention in Transformer is not sensitive to short-term feature dependence, which leads to model anomalies in time-series; (2) Memory bottleneck: The spatial complexity of regular transformation increases twice with the sequence length, making direct modeling of long time-series infeasible. In order to solve these problems, this paper designs an efficient model for long time-series prediction. It is a double pyramid bidirectional feature fusion mechanism network with parallel Temporal Convolution Network (TCN) and FastFormer. This network structure can combine the time series fine-grained information captured by the Temporal Convolution Network with the global interactive information captured by FastFormer, it can well handle the time series prediction problem.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2022EAP1081/_p

Copy

@ARTICLE{e106-a_6_886,
author={Na WANG, Xianglian ZHAO, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Time-Series Prediction Based on Double Pyramid Bidirectional Feature Fusion Mechanism},
year={2023},
volume={E106-A},
number={6},
pages={886-895},
abstract={The application of time-series prediction is very extensive, and it is an important problem across many fields, such as stock prediction, sales prediction, and loan prediction and so on, which play a great value in production and life. It requires that the model can effectively capture the long-term feature dependence between the output and input. Recent studies show that Transformer can improve the prediction ability of time-series. However, Transformer has some problems that make it unable to be directly applied to time-series prediction, such as: (1) Local agnosticism: Self-attention in Transformer is not sensitive to short-term feature dependence, which leads to model anomalies in time-series; (2) Memory bottleneck: The spatial complexity of regular transformation increases twice with the sequence length, making direct modeling of long time-series infeasible. In order to solve these problems, this paper designs an efficient model for long time-series prediction. It is a double pyramid bidirectional feature fusion mechanism network with parallel Temporal Convolution Network (TCN) and FastFormer. This network structure can combine the time series fine-grained information captured by the Temporal Convolution Network with the global interactive information captured by FastFormer, it can well handle the time series prediction problem.},
keywords={},
doi={10.1587/transfun.2022EAP1081},
ISSN={1745-1337},
month={June},}

Copy

TY - JOUR
TI - Time-Series Prediction Based on Double Pyramid Bidirectional Feature Fusion Mechanism
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 886
EP - 895
AU - Na WANG
AU - Xianglian ZHAO
PY - 2023
DO - 10.1587/transfun.2022EAP1081
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E106-A
IS - 6
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - June 2023
AB - The application of time-series prediction is very extensive, and it is an important problem across many fields, such as stock prediction, sales prediction, and loan prediction and so on, which play a great value in production and life. It requires that the model can effectively capture the long-term feature dependence between the output and input. Recent studies show that Transformer can improve the prediction ability of time-series. However, Transformer has some problems that make it unable to be directly applied to time-series prediction, such as: (1) Local agnosticism: Self-attention in Transformer is not sensitive to short-term feature dependence, which leads to model anomalies in time-series; (2) Memory bottleneck: The spatial complexity of regular transformation increases twice with the sequence length, making direct modeling of long time-series infeasible. In order to solve these problems, this paper designs an efficient model for long time-series prediction. It is a double pyramid bidirectional feature fusion mechanism network with parallel Temporal Convolution Network (TCN) and FastFormer. This network structure can combine the time series fine-grained information captured by the Temporal Convolution Network with the global interactive information captured by FastFormer, it can well handle the time series prediction problem.
ER -