Chege-N/Kenya-Coffee-ML-Analysis

Interactive dashboard for the Kenya coffee sector (2018–2020) using 2,784 growers and 157 dealers. Includes ML models (Random Forest, KMeans, anomaly detection), cluster profiles, geo-visualisation, 12-month and 20 year-registration projections. Built with vanilla HTML/CSS + Chart.js. # Kenya Coffee Sector — Advanced ML Analysis Project **Author:** Chege-N | **Dataset:** Kenya Coffee Growers (2,784) + Dealers (157) | **Period:** 2018–2020 --- ## Project Structure ``` kenya_coffee_project/ ├── data/ │ ├── coffeedealers_raw.csv # Original dealer data (157 records) │ ├── coffeegrowers_raw.csv # Original grower data (2,784 records) │ ├── growers_processed.csv # Feature-engineered growers │ ├── dealers_processed.csv # Feature-engineered dealers │ └── geo_data.json # 897 georeferenced growers (JSON) ├── analysis/ │ └── kenya_coffee_analysis.py # Full ML pipeline script ├── models/ │ └── (model artifacts if saved) └── outputs/ ├── kenya_coffee_dashboard.html # Interactive dashboard (open in browser) └── analysis_summary.json # Machine-readable results summary ``` --- ## Datasets ### Coffee Growers (`coffeegrowers.csv`) - **2,784 records**, 25 columns - Date range: 2018-04-28 → 2020-09-18 - Key fields: `id`, `title`, `actor`, `lat`, `lon`, `county_name_id`, `active`, `created`, `updated` - Actor types: Factory (983), Small Estates (984), Cooperative Society (494), Estate Producers (320) ### Coffee Dealers (`coffeedealers.csv`) - **157 records**, 7 columns - Key fields: `id`, `title`, `license_number`, `nce_ref`, `notes`, `website` --- ## ML Models Applied | Model | Purpose | Key Metric | |-------|---------|-----------| | **Random Forest** (n=200) | Predict grower active status | F1=0.630 (5-fold CV) | | **Gradient Boosting** | Compare classifier | F1=0.580 | | **Logistic Regression** | Baseline classifier | F1=0.410 | | **KMeans** (k=4) | Grower segmentation | 4 distinct clusters | | **Isolation Forest** | Anomaly detection | 136 anomalies (4.9%) | | **PCA** (2D) | Dimensionality reduction | 68.2% variance explained | | **Linear Regression** | Registration trend + projection | R²=0.21 | --- ## Key Findings ### Growers 1. **Only 10.8% of growers are active** (301/2,784). Sm …