RESAMPLING AND ENSEMBLE LEARNING TO IMPLOVE ACCURACY OF DEEP LEARNING RAINFALL-RUNOFF MODELING / リサンプリングとアンサンブル学習を用いた深層学習降雨流出モデルの精度向上の試み
- Resource Type
- Journal Article
- Authors
- Daiju SAKAGUCHI; Kei ISHIDA; Takeyoshi NAGASATO; 坂口 大珠; 永里 赳義; 石田 桂
- Source
- Intelligence, Informatics and Infrastructure / AI・データサイエンス論文集. 2022, 3(J2):906
- Subject
- Deep learning
Ensemble learning
Long Short-Term Memory
Machine learning
Rainfallrunoff modling
Resampling
River flow estimation
XGBoost
- Language
- Japanese
- ISSN
- 2435-9262
This study used resampling and ensemble learning. This study tried to improve the accuracy of rainfall runoff modeling. Resampling was applied to rainfall-runoff modeling using LSTM. Resampling focused on river flow, which should be focused on River flow was estimated by focused learning. Ensemble learning tried to improve the estimation accuracy of the entire river flow rate using XGBoost. The input data was an estimated flow rate that captured different characteristics with LSTM. As a result, the river flow range that we want to focus on is learned by resampling. This has been shown to contribute to improving the accuracy of river flow estimation. In ensemble learning, the estimated river flow rate that captures different characteristics is input. It was shown that this does not improve the accuracy in rainfall runoff modeling.