TITLE:
Application of Machine Learning for Flood Prediction and Evaluation in Southern Nigeria
AUTHORS:
Emeka Bright Ogbuene, Chukwumeuche Ambrose Eze, Obianuju Getrude Aloh, Andrew Monday Oroke, Damian Onuora Udegbunam, Josiah Chukwuemeka Ogbuka, Fred Emeka Achoru, Vivian Amarachi Ozorme, Obianuju Anwara, Ikechukwu Chukwunonyelum, Anthonia Nneka Nebo, Obiageli Jacinta Okolo
KEYWORDS:
Machine Learning, Flood, Prediction, Evaluation, Southern Nigeria
JOURNAL NAME:
Atmospheric and Climate Sciences,
Vol.14 No.3,
July
3,
2024
ABSTRACT: This study explored the application of machine learning techniques for flood prediction and analysis in southern Nigeria. Machine learning is an artificial intelligence technique that uses computer-based instructions to analyze and transform data into useful information to enable systems to make predictions. Traditional methods of flood prediction and analysis often fall short of providing accurate and timely information for effective disaster management. More so, numerical forecasting of flood disasters in the 19th century is not very accurate due to its inability to simplify complex atmospheric dynamics into simple equations. Here, we used Machine learning (ML) techniques including Random Forest (RF), Logistic Regression (LR), Naïve Bayes (NB), Support Vector Machine (SVM), and Neural Networks (NN) to model the complex physical processes that cause floods. The dataset contains 59 cases with the goal feature “Event-Type”, including 39 cases of floods and 20 cases of flood/rainstorms. Based on comparison of assessment metrics from models created using historical records, the result shows that NB performed better than all other techniques, followed by RF. The developed model can be used to predict the frequency of flood incidents. The majority of flood scenarios demonstrate that the event poses a significant risk to people’s lives. Therefore, each of the emergency response elements requires adequate knowledge of the flood incidences, continuous early warning service and accurate prediction model. This study can expand knowledge and research on flood predictive modeling in vulnerable areas to inform effective and sustainable contingency planning, policy, and management actions on flood disaster incidents, especially in other technologically underdeveloped settings.