Optimizing Flood Hazard Prediction Parameters Using LSTM and GIS: A Case Study of Tetouan City, Morocco

The problem of climate change negatively affects the world, causing several devastating natural catastrophes among them floods. Their consequences are significant for human lives and materials. Particularly in Morocco frequent and severe floods have taken place in various territories, putting the country it in great need of proactive modeling and management to avoid such a risk. Different methods have been employed for this purpose especially Artificial Intelligent (AT) and Geographic Information Systems (GTS). Therefore, this work aims to enhance the accuracy of deep learning techniques through extensive experimentation and testing of different optimization strategies in Tetouan in the north of Morocco. The dataset includes 11 flood conditioning factors, along with the various optimization techniques to augment the accuracy of the Long Short-Term Memory (LSTM) model, such as batch normalization, data augmentation, and Information Gain Ratio (IGR). After the analysis, the results show that the LSTM model achieved a high accuracy of 95.21%.