gabrielntwari/Clean-Cooking-Adoption-Analysis

Machine learning analysis of clean cooking adoption patterns in Rwanda. Uses Random Forest, Neural Networks, and geospatial analysis to identify intervention hotspots and predict adoption barriers for targeted support programs. # DelAgua Clean Cooking Analytics **Machine Learning Analysis of Clean Cooking Adoption Patterns in Rwanda** This project analyzes clean cooking stove adoption barriers in Rwanda using advanced machine learning techniques and geospatial analysis to identify intervention hotspots for targeted support programs. ## 📋 Overview This analysis was conducted as part of a data analyst practical assessment for **DelAgua**, focusing on understanding the factors that influence clean cooking technology adoption in Rwandan communities. The study combines household survey data with geospatial information to predict adoption patterns and map high-priority intervention zones. ## 🎯 Objectives - **Identify key barriers** to clean cooking stove adoption - **Predict low-adoption areas** using machine learning models - **Map geospatial hotspots** for targeted interventions - **Provide actionable insights** for program implementation ## 🔬 Methodology ### Data Analysis Pipeline 1. **Exploratory Data Analysis** - Understanding adoption patterns and demographics 2. **Feature Engineering** - Creating predictive variables from survey responses 3. **Machine Learning Modeling** - Multiple algorithms for adoption prediction 4. **Geospatial Mapping** - Hotspot identification and visualization ### Models Implemented - **Random Forest Classifier** - Ensemble learning for feature importance - **Gradient Boosting** - Advanced tree-based prediction - **Neural Networks (TensorFlow/Keras)** - Deep learning approach - **Logistic Regression** - Baseline comparison model ### Key Features Analyzed - Household demographics and income - Geographic location (longitude/latitude) - Access to markets and infrastructure - Family size and composition - Regional characteristics ## 📊 Key Findings ### High-Risk Intervention Zones Identified: 1. **Western Peak** (Longitude ~29.5, Latitude ~-1.6) - Highest concentration of low adoption - Geographically isolated areas - Priority zone for intervention 2. **No …