Predicting climate induced floods using machine learning

Abstract

Climate change has been causing devastation on the economy of the country and the world as a whole. The study aims to determine how climate change would impact the frequency and severity of one of the major natural disasters which is flood. Data sets containing information about the rise of global temperatures, annual rainfall, sea level rise and flood occurrences were surveyed and assembled. Different attributes from the assembled datasets were then taken out and spliced into datasets that suit the scope of our research. A plethora of machine learning algorithms have been used to develop different prediction models based on the constructed datasets. Algorithms employed in the development of flood prediction models include: “Logistic Regression”, “Decision Tree”, “K Nearest Neighbors”, “Support Vector Machine”, “Random Forest” and “Ensemble Learning”. Projection models were then trained by employing an “Autoregressive” approach for generating projection data, which were a prerequisite for the flood prediction models in making predictions of future flood incidents. And with the aid of the generated projection data, predictions of flood incidents were made for the years starting from 2022 to 2050.