Ijegwa et al. (2025) Intelligent Web App for Flash Flood Prediction in Nigeria’s Coastal Regions

Identification

• Journal: ABUAD Journal of Engineering Research and Development (AJERD)
• Year: 2025
• Date: 2025-10-19
• Authors: Acheme David Ijegwa, Mohammed Abduljalal
• DOI: 10.53982/ajerd.2025.0803.07-j

Research Groups

• Department of Computer Science, College of Science and Computing, Wigwe University, Isiokpo
• Department of Computer Science, Edo State University, Uzairue

Short Summary

This study developed an intelligent web application using a Random Forest machine learning model to predict flash flood occurrences in Nigeria's coastal regions, achieving 96% accuracy and real-time latency of less than one second.

Objective

• To develop an intelligent web application for accurate, real-time flash flood prediction in Nigeria's coastal regions using machine learning, providing timely forecasts and recommendations to aid disaster preparedness and response.

Study Configuration

• Spatial Scale: Nigeria's coastal regions (e.g., Lagos, Bayelsa, Rivers, Cross River, and Delta States).
• Temporal Scale: Historical data covered five years (2018–2023); the system provides real-time predictions with latency less than one second.

Methodology and Data

• Models used:
  • Machine Learning: Random Forest (primary), Decision Tree, Logistic Regression, Support Vector Classifier (for comparative analysis).
  • Web Development: Python (FLASK framework), Scikit-learn, TypeScript, React.js, Tree.js, MongoDB.
• Data sources:
  • Real-time environmental data (rainfall intensity, river levels, soil moisture, temperature, humidity, wind speed, visibility) captured via the OpenMeteo API.
  • Historical weather data from the OpenMeteo API (15,320 records covering 2018–2023).

Main Results

• The Random Forest model achieved an accuracy of 96%, precision of 75%, recall of 91%, and an F1-score of 82.2%.
• The system demonstrated a real-time latency of less than one second, indicating a fast response to changing environmental data.
• Comparative analysis showed Random Forest outperformed other models: Decision Tree (Accuracy 91.2%), Logistic Regression (Accuracy 89.8%), and Support Vector Classifier (Accuracy 87.6%).
• Strong positive correlations were observed between rainfall and river discharge, while negative correlations were noted between temperature and soil moisture, confirming the relevance of selected features.

Contributions

• Integration of real-time environmental variables obtained from the OpenMeteo API with machine learning algorithms, ensuring dynamic adaptation to changing weather and environmental conditions.
• Comprehensive comparative analysis of Random Forest, Decision Tree, Logistic Regression, and Support Vector Classifier, justifying Random Forest as the optimal algorithm for this application.
• Successful deployment of the trained model into a practical web application with an intuitive and user-friendly frontend and a scalable backend, bridging the gap between advanced flood prediction models and real-world accessibility for at-risk communities.

Funding

• Not specified in the provided paper text.

Citation

@article{Ijegwa2025Intelligent,
  author = {Ijegwa, Acheme David and Abduljalal, Mohammed},
  title = {Intelligent Web App for Flash Flood Prediction in Nigeria’s Coastal Regions},
  journal = {ABUAD Journal of Engineering Research and Development (AJERD)},
  year = {2025},
  doi = {10.53982/ajerd.2025.0803.07-j},
  url = {https://doi.org/10.53982/ajerd.2025.0803.07-j}
}