Prediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU model
Abstract Accurate tide level prediction is crucial to human activities in coastal areas. Many practical applications show that compared with traditional harmonic analysis, long short-term memory (LSTM), gated recurrent units (GRUs) and other neural networks, along with ensemble learning models, such...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-01-01
|
| Series: | Scientific Reports |
| Online Access: | https://doi.org/10.1038/s41598-022-26213-y |
| _version_ | 1797958650152615936 |
|---|---|
| author | Ye Su Xuchu Jiang |
| author_facet | Ye Su Xuchu Jiang |
| author_sort | Ye Su |
| collection | DOAJ |
| description | Abstract Accurate tide level prediction is crucial to human activities in coastal areas. Many practical applications show that compared with traditional harmonic analysis, long short-term memory (LSTM), gated recurrent units (GRUs) and other neural networks, along with ensemble learning models, such as light gradient boosting machine (LightGBM) and eXtreme gradient boosting (XGBoost), can achieve extremely high prediction accuracy in relatively stationary time series. Therefore, this paper proposes a variable weight combination model based on LightGBM and CNN-BiGRU with relevant research. It uses the variable weight combination method to weight and synthesize the prediction results of the two base models so that the combination model has a stronger ability to capture time series features and fits the data well. The experimental results show that in contrast to the base model LightGBM, the RMSE value and MAE value of the combination model are reduced by 43.2% and 44.7%, respectively; in contrast to the base model CNN-BiGRU, the RMSE value and MAE value of the combination model are reduced by 35.3% and 39.1%, respectively. This means that the variable weight combination model can greatly improve the accuracy of tide level prediction. In addition, we use tidal data from different geographical environments to further verify the good universality of the model. This study provides a new idea and method for tide prediction. |
| first_indexed | 2024-04-11T00:21:58Z |
| format | Article |
| id | doaj.art-7f2729c6ad264d2db4c981ed4245c56d |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-04-11T00:21:58Z |
| publishDate | 2023-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-7f2729c6ad264d2db4c981ed4245c56d2023-01-08T12:12:42ZengNature PortfolioScientific Reports2045-23222023-01-0113111310.1038/s41598-022-26213-yPrediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU modelYe Su0Xuchu Jiang1School of Statistics and Mathematics, Zhongnan University of Economics and LawSchool of Statistics and Mathematics, Zhongnan University of Economics and LawAbstract Accurate tide level prediction is crucial to human activities in coastal areas. Many practical applications show that compared with traditional harmonic analysis, long short-term memory (LSTM), gated recurrent units (GRUs) and other neural networks, along with ensemble learning models, such as light gradient boosting machine (LightGBM) and eXtreme gradient boosting (XGBoost), can achieve extremely high prediction accuracy in relatively stationary time series. Therefore, this paper proposes a variable weight combination model based on LightGBM and CNN-BiGRU with relevant research. It uses the variable weight combination method to weight and synthesize the prediction results of the two base models so that the combination model has a stronger ability to capture time series features and fits the data well. The experimental results show that in contrast to the base model LightGBM, the RMSE value and MAE value of the combination model are reduced by 43.2% and 44.7%, respectively; in contrast to the base model CNN-BiGRU, the RMSE value and MAE value of the combination model are reduced by 35.3% and 39.1%, respectively. This means that the variable weight combination model can greatly improve the accuracy of tide level prediction. In addition, we use tidal data from different geographical environments to further verify the good universality of the model. This study provides a new idea and method for tide prediction.https://doi.org/10.1038/s41598-022-26213-y |
| spellingShingle | Ye Su Xuchu Jiang Prediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU model Scientific Reports |
| title | Prediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU model |
| title_full | Prediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU model |
| title_fullStr | Prediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU model |
| title_full_unstemmed | Prediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU model |
| title_short | Prediction of tide level based on variable weight combination of LightGBM and CNN-BiGRU model |
| title_sort | prediction of tide level based on variable weight combination of lightgbm and cnn bigru model |
| url | https://doi.org/10.1038/s41598-022-26213-y |
| work_keys_str_mv | AT yesu predictionoftidelevelbasedonvariableweightcombinationoflightgbmandcnnbigrumodel AT xuchujiang predictionoftidelevelbasedonvariableweightcombinationoflightgbmandcnnbigrumodel |