NsabimanaSteve/burundi-climate-forecasting

Machine learning pipeline for climate and hydrological forecasting in Burundi , precipitation, temperature, Rusizi River and Lake Tanganyika water levels (2025–2030) # burundi-climate-forecasting Machine learning pipeline for climate and hydrological forecasting in Burundi , precipitation, temperature, Rusizi River and Lake Tanganyika water levels (2025–2030) PROJECT TITLE: EXPLORING ARTIFICIAL INTELLIGENCE FOR ENHANCING CLIMATE CHANGE FORECASTING AND MITIGATION STRATEGIES IN BURUNDI AUTHOR: Eunice Sayubu & Steve Nsabimana ID: 60192026 & 36422026 --- 1. GITHUB REPOSITORY The full source code for this project is available at: github.com --- 2. PROJECT OVERVIEW This project investigates the use of Artificial Intelligence models to improve climate forecasting and disaster risk mitigation in Burundi, a country highly vulnerable to climate change but lacking localized AI-based systems. Multiple models, including Linear Regression, ARIMA, Random Forest, XGBoost, LSTM, CNN, and ensemble methods, were implemented and compared across four climate variables: precipitation, temperature, and water levels of the Rusizi River and Lake Tanganyika. The results show that no single model consistently outperformed others, with different models excelling depending on the variable being predicted. Traditional statistical methods such as ARIMA remained competitive, outperforming AI models in certain cases like Lake Tanganyika water level prediction. Overall, the study demonstrates the potential of combining AI and traditional approaches to build reliable, data-driven climate forecasting systems in data-scarce environments like Burundi. It includes: * Data preprocessing and cleaning * Feature selection and scaling * Training multiple models (e.g., Linear Regression, Random Forest, XGBoost, LSTM) * Model evaluation using standard metrics * Model interpretation using SHAP and LIME * Future Prediction (2025-2030) --- 3. SYSTEM REQUIREMENTS To run this project, ensure the following are installed: * Python 3.8 or higher * pip (Python package manager) Required Python libraries include: num …