Deep Learning for Tidal Flood Prediction in West Pandeglang Waters, Banten

Abstract

Tidal flooding poses a significant threat to coastal areas, exacerbated by rising sea levels. In West Pandeglang Waters, Banten, frequent tidal floods impact communities, necessitating accurate prediction models for effective disaster mitigation. This study aims to develop a deep learning-based tidal flood prediction model using Keras and TensorFlow. The model incorporates oceanic and atmospheric variables, including sea surface height, wave characteristics, wind components, and precipitation data from 2003 to 2023. To address data imbalance, Synthetic Minority Over-sampling Technique (SMOTE) and MinMax scaling were applied, ensuring balanced class distribution. The model was trained and evaluated using a dataset comprising 11,808 samples, achieving an accuracy of 86% and an area under the curve (AUC) of 0.93. These results indicate a strong capability to differentiate between flood and non-flood conditions. The study demonstrates the effectiveness of deep learning in predicting tidal floods, providing valuable insights for early warning systems and coastal management in flood-prone regions.