ABSTRACT
Flooding remains a critical environmental challenge in Benin City, Nigeria, exacerbated by
rapid urbanization, inadequate drainage systems, and climate change. This project proposes an
Artificial Intelligence (AI)-based Flood Early Warning System (FEWS) to enhance flood risk
prediction and preparedness. Leveraging meteorological data from the Nigerian
Meteorological Agency (NiMet)—including rainfall, temperature, relative humidity, and
evapotranspiration—the system employs a Random Forest machine learning model to classify
flood risks into low and high categories. A user-friendly web interface, built using Streamlit
and hosted on Hugging Face Spaces, allows stakeholders to input real-time weather data and
receive instant flood risk assessments. Additionally, the system integrates BulkSMSNigeria
API to send automated SMS alerts when high flood risks are detected, ensuring rapid
dissemination of warnings.
Through rigorous testing, the system demonstrated high accuracy in flood prediction, effective
cloud deployment, and reliable SMS alert functionality. Data preprocessing techniques, such
as SMOTE, addressed class imbalances in the dataset, further enhancing model performance.
This AI-powered FEWS represents a significant advancement in flood management, offering
a cost-effective, scalable solution for disaster risk reduction in Benin City and similar flood
prone areas.
