Research Article
Data-Driven Fast Real-Time Flood Forecasting Model for Processing Concept Drift
@INPROCEEDINGS{10.1007/978-3-030-48513-9_30, author={Le Yan and Jun Feng and Yirui Wu and Tingting Hang}, title={Data-Driven Fast Real-Time Flood Forecasting Model for Processing Concept Drift}, proceedings={Cloud Computing, Smart Grid and Innovative Frontiers in Telecommunications. 9th EAI International Conference, CloudComp 2019, and 4th EAI International Conference, SmartGIFT 2019, Beijing, China, December 4-5, 2019, and December 21-22, 2019}, proceedings_a={CLOUDCOMP}, year={2020}, month={6}, keywords={Medium and small watershed Concept drift Data-driven model Fast real-time flood forecasting}, doi={10.1007/978-3-030-48513-9_30} }
- Le Yan
Jun Feng
Yirui Wu
Tingting Hang
Year: 2020
Data-Driven Fast Real-Time Flood Forecasting Model for Processing Concept Drift
CLOUDCOMP
Springer
DOI: 10.1007/978-3-030-48513-9_30
Abstract
The hydrological data of small and medium watershed develops with the passage of time. The rainfall-runoff patterns in these data often develop over time, and the models established for the analysis of such data will soon not be applicable. In view of the problem that adaptability and accuracy of the existing data-driven flood real-time forecasting model in medium and small watershed with concept drift. We update the data-driven model using incremental training based on support vector machine (SVM) and gated recurrent unit (GRU) model respectively. According to the rapid real-time flood forecasting test results of the Tunxi watershed, Anhui Province, China, the fast real-time flood forecast data-driven model with incremental update can more accurately predict the moment when the flood begins to rise and the highest point of flood stream-flow, and it is an effective tool for real-time flood forecasting in small and medium watersheds.