The increasing frequency of flood events worldwide is a significant concern, particularly for the United Nations (UN), as it impacts both economic stability and public safety. In recent decades, floods have caused extensive damage to lives and properties, and projections indicate this trend will continue. Rapid evacuation in the event of a flood relies on effective early warning systems that can predict floods and encourage residents to evacuate high-risk areas. Africa is the third most affected continent by floods, following Asia and Europe. However, the development and implementation of flood forecasting models in Africa are still in the early stages. The continent has only a few hydrological models for flood forecasting, and most of these need improvements to keep pace with changing conditions. The study aimed to enhance flood prediction and advisory systems by leveraging machine learning (ML). The objectives are to develop a predictive model, create dual subsystems for web and Android applications, disseminate predictions and advisories through multiple communication channels, and incorporate persuasive messages to prompt at-risk populations to evacuate. A survey was conducted in Cross River State to evaluate the impact of these persuasive messages and the overall support for the Flood Predictive and Advisory System (FPAS). The study utilized the NiMet dataset, preprocessed the data, and trained the model using SVM, Random Forest, and XGBoost algorithms. Subsequently, both the FPAS web and Android applications were developed and rigorously tested. The research methodology applied in this work was a hybrid approach, combining Object Oriented Analysis and Design Methodology (OOADM) and the Cross Industry Standard Process for Data Mining (CRISP-DM). OOADM was used to develop the mobile app for citizens’ registration and a web app for administrative activities, while CRISP-DM was used to create a data-drive predictive model for the app. These applications facilitated citizen registration and administrative activities, providing feedback via websites, emails, and automated alerts. The survey results revealed that 76.56% of respondents had not experienced persuasive messages before, while 91.15% expressed strong support for the FPAS system incorporating these techniques. Finally, the study successfully developed an advanced flood prediction model and an integrated advisory system, demonstrating significant potential in enhancing disaster preparedness and response through innovative technology and strategic communication.
Search from 250 M+ research papersSearch from 250 M+ research papers
Jun 30, 2024
Peter Tawor Etta + 2
Open Access