Answers are grounded in this paper and cite their sources.
Flooding is a natural disaster with devastating socio-economic and environmental impacts globally. The existing methods rely on the incomplete or sparse historical data, leading to less reliable predictions. Also, deals with a narrow set of features, missing out on critical aspects of rainfall and hydrological dynamics which affects its generalisability. To address these challenges, this paper presents a novel approach, the Hybrid Deep Architecture with Entropy-based Flood Prediction (HDAE-FP) model for rainfall-based flood prediction. The process begins with data acquisition, followed by preprocessing to normalise the data by using ANAMM scaling Normalization for uniformity across features. Feature extraction entails computing statistical features such as mean, median, standard deviation, variance, skewness, kurtosis, and percentiles, alongside calculating MKE-based entropy measures to capture rainfall pattern uncertainty and hydrological dynamics complexity. These extracted features serve as input variables for prediction models. For the rainfall and flood prediction, a hybrid architecture is adopted that combines both Modified Attention-based Long Short-Term Memory (MA-LSTM) and Bi-GRU. As a result, the proposed MA-LSTM+Bi-GRU model has attained an accuracy of 0.957 at the 90% of training which surpasses the outcomes of the prediction compared to other current techniques.
No takes yet. Share an insight, caveat, or question.
Yoganand et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: