AIBullisharXiv – CS AI · 8h ago6/10
🧠
Physics-Informed Machine Learning for Short-Term Flood Prediction
Researchers propose a Physics-Informed Machine Learning framework that integrates hydrological constraints into LSTM neural networks to improve flood prediction accuracy in data-scarce environments. The approach demonstrates superior performance over standard deep learning models, particularly during extreme weather events, by enforcing physically plausible behavior through a Trend Alignment constraint in the loss function.