Prediction of cold region dew volume based on an ECOA-BiTCN-BiLSTM hybrid model

This paper presents a hybrid prediction model, ECOA-BiTCN-BiLSTM, for predicting dew in cold areas. The model integrates BiTCN and BiLSTM neural networks to enhance performance. An enhanced Crayfish optimization algorithm (ECOA) with four mixed strategies was employed to optimize the model’s hyperpa...

Full description

Saved in:
Bibliographic Details
Published in:Scientific reports Vol. 15; no. 1; pp. 4265 - 14
Main Authors: Zhang, Yi, Liu, Pengtao, Xu, Yingying, Zhang, Meng
Format: Journal Article
Language:English
Published: London Nature Publishing Group UK 04.02.2025
Nature Publishing Group
Nature Portfolio
Subjects:
639/705
704/106
704/172
Accuracy
Agricultural land
Agricultural production
Algorithms
Artificial intelligence
BiTCN-BiLSTM model
Cold
Cold regions
Deep learning
Dew
Dew volume prediction
Enhanced crayfish optimization algorithm
Genetic algorithms
Humanities and Social Sciences
Humidity
Hyperparameter optimization
Machine learning
multidisciplinary
Neural networks
Optimization algorithms
Prediction models
Science
Science (multidisciplinary)
Support vector machines
Swarm intelligence optimization algorithm
Hyperparameter optimization
Dew volume prediction
BiTCN-BiLSTM model
Swarm intelligence optimization algorithm
Enhanced crayfish optimization algorithm
ISSN:2045-2322, 2045-2322
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper presents a hybrid prediction model, ECOA-BiTCN-BiLSTM, for predicting dew in cold areas. The model integrates BiTCN and BiLSTM neural networks to enhance performance. An enhanced Crayfish optimization algorithm (ECOA) with four mixed strategies was employed to optimize the model’s hyperparameters and reduce the impact of arbitrary selection. The proposed ECOA-BiTCN-BiLSTM model was validated using dew data from farmland in a northeastern Chinese city. Comparative experiments were conducted against the BiTCN model, the BiLSTM model, the original BiTCN-BiLSTM model, and other models optimized with advanced swarm intelligence algorithms. The experimental results demonstrate that the proposed model achieved a mean absolute error (MAE) of 0.002424, a root mean square error (RMSE) of 0.003984, and a mean absolute percentage error (MAPE) of 0.123050, with a coefficient of determination R 2 of 0.999840. These results indicate that the ECOA-BiTCN-BiLSTM model outperforms the other prediction models across all evaluated metrics, offering higher prediction accuracy and highly effective prediction models.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:2045-2322
2045-2322
DOI:10.1038/s41598-024-74097-x