Publication

Abstract : Natural disasters such as floods are some of the most devastating occurrences, causing wide-scale damage to infrastructure, loss of lives, and widespread social disruption. Conventional flood forecasting techniques are unreliable and do not provide sufficient warning to stakeholders of such occurrences. In this paper, we present the use of machine learning (ML) algorithms in predicting flood damage, citing implications of magnitude, economic loss, casualties, and displacement. From historical meteorological and hydrological records, we compare the performance of Linear Regression, Random Forest, XGBoost, and Support Vector Machines (SVM) using different approaches. The results show that SVM is the best tool to classify flood severity, while Random Forest and XGBoost provide a better performance in predicting the non-linear relations typical of complex flooding damage. The conclusions from the research are that ML models can improve the accuracy of flood damage prediction, promote improved decision-making and planning, and mitigate human and economic losses from floods.