Ajzelly/kenya-labour-market-ml-dashboard

ML-powered Kenya unemployment rate predictor with Power BI dashboard # 🇰🇪 Kenya Labour Market Intelligence Dashboard > End-to-end Machine Learning project predicting Kenya's > unemployment rate using World Bank economic indicators (1991–2024), > with an interactive Power BI dashboard. --- ## Dashboard Preview ### Page 1 — Economic Overview ### Page 2 — ML Model Results ### Page 3 — What Drives Unemployment --- ## Machine Learning Results | Metric | Score | |--------|-------| | Algorithm | Linear Regression | | R² Score | 0.994 | | MAE | 0.059 percentage points | | Training years | 1991 – 2024 (33 data points) | | Best vs | Outperformed Random Forest & Gradient Boosting | --- ## Key Findings - **Labour Force Participation** is the strongest predictor of Kenya's unemployment rate (importance score: 0.64) - **Youth Unemployment** acts as an early warning signal for total unemployment (importance score: 0.33) - **COVID-19 (2020)** caused the largest single-year spike — unemployment jumped from 2.8% to 5.7% - **Linear Regression outperformed** complex models — with only 33 data points, simpler models generalise better --- ## Project Structure ``` ├── data/ # World Bank raw CSVs ├── output/ # ML model exports for Power BI │ ├── kenya_unemployment_predictions.csv │ └── kenya_feature_importance.csv ├── screenshots/ # Power BI dashboard pages ├── kenya_labour_market.ipynb # Full Colab notebook ├── Kenya_Labour_Market_Dashboard.pbix # Power BI file └── README.md ``` --- ## Tools & Technologies | Category | Tools | |----------|-------| | Language | Python 3 | | Data wrangling | Pandas, NumPy | | Machine Learning | Scikit-learn | | Visualisation | Matplotlib, Seaborn | | Dashboard | Microsoft Power BI | | Environment | Google Colab | | Data source | World Bank Open Data | --- ## ML Pipeline 1. **Data Collection** — 8 World Bank indicators for Kenya 2. **Data Cleaning** — Forward fill missing values, filter 1991–2024 3. **Feature Engineering** — L …