MarialRK/interpretable-flood-prediction

Interpretable Flood-Risk Prediction System for the Sudd Wetland Region, South Sudan # 🌊 Interpretable Flood-Risk Prediction System ### Sudd Wetland Region, South Sudan An interpretable machine learning system for predicting flood risk in the Sudd Wetland Region of South Sudan using satellite-derived environmental data and historical flood information. The project combines multiple machine learning models with explainable AI (SHAP) to support disaster preparedness and informed decision-making. --- ## 📖 Overview Flooding is one of the most destructive natural hazards affecting communities across South Sudan. This project develops an interpretable flood prediction system capable of providing **14-day early warning forecasts** by integrating multiple environmental datasets and comparing six machine learning models. The application provides: - 🌧️ Flood risk prediction - 📊 Interactive visualizations - 🧠 Explainable AI using SHAP - 📈 Model comparison dashboard - 🚨 Actionable recommendations for disaster preparedness --- ## ✨ Key Features - Predicts flood risk using six machine learning models - Compares classical ML and deep learning approaches - Uses SHAP explainability for transparent predictions - Interactive Streamlit dashboard - 14-day flood forecasting capability - Built specifically for the Sudd Wetland Region of South Sudan # 📸 Application Screenshots ## screenshot_high_risk.png ## screenshot_low_risk.png ## screenshot_shap_analysis.png ## screenshot_model_comparison.png ## 🌐 Live Application 🔗 **Streamlit App** sudd-flood-prediction.strea… --- ## 📹 Project Demonstration 🎥 **Video Walkthrough** screenrec.com --- # 🏆 Model Performance | Model | Accuracy | Precision | Recall | F1-Score | ROC-AUC | |-------|----------:|----------:|--------:|----------:|---------:| | **Support Vector Machine (Best)** | **68.06%** | **64.29%** | **62.70%** | **63.49%** | **70.81%** | | ARIMA | 66.67% | 61.54% | 58.82% | 60.00% | 64.83% | | XGBoost | 61.11% | 57.14% | 50.00% | 53.33% | 59.69% …