Advanced Machine Learning Techniques for Predicting Coastal Line Changes in Northern Coast of Java
Atik Nurul Aini (a*)

(a) Cerdas Antisipasi Risiko bencana Indonesia (CARI!), Jalan Sepak Bola 5, Bandung 40293, Indonesia

Abstract

Coastal erosion and the shifting of coastal lines bear significant implications for the management and preservation of the northern coast of Java. This study employs sophisticated machine learning techniques to forecast variations in coastal lines, with the intention of advancing informative coastal management and mitigating associated risks. The methodology involves the integration of Landsat satellite image analysis and pertinent coastal line reanalysis data. Within this framework, several machine learning models are embraced, encompassing the Neural Network Auto-Regression (NNAR), Long Short-Term Memory (LSTM), eXtreme Gradient Boosting (XGBoost), Light Gradient-Boosting Machine (LGBM), Artificial Neural Network (ANN), k-Nearest Neighbor (KNN), and Support Vector Machine (SVM). Through meticulous comparison of these machine learning models, we elucidate their superior predictive accuracy, particularly under extreme weather conditions. Our research findings underscore the efficacy of these models in predicting both short-term and long-term changes in the coastal lines of the northern coast of Java, while accentuating their potential to fortify coastal protection strategies. In-depth analysis unveils that wind speed and wave height emerge as the most pivotal features in predicting coastal line shifts. The holistic performance of the models is scrutinized, and the observational results indicate the outperformance of the ANN model compared to the remaining six models within this context.

Keywords: coastal line changes- machine learning- NNAR- LSTM- XGBoost- LGBM- ANN- KNN- SVM

Topic: Marine Hazard, and Coastal Protection