Upload folder using huggingface_hub

README.md

@@ -0,0 +1,17 @@
+---
+license: cc-by-4.0
+task_categories:
+  - tabular-classification
+  - tabular-regression
+language:
+  - en
+tags:
+  - poverty
+  - social-protection
+  - africa
+  - synthetic-data
+  - sub-saharan-africa
+  - poverty-headcount
+size_categories:
+  - 10K<n<100K
+---

generate_dataset.py

@@ -0,0 +1,126 @@
+"""
+Poverty Headcount Africa Dataset Generator
+
+Parameter Evidence Table:
+| Parameter | Source | Value | Notes |
+|-----------|--------|-------|-------|
+| $2.15/day poverty line | World Bank | $2.15 | Extreme poverty threshold |
+| $3.00/day | World Bank 2024 | 46% of SSA | Current poverty rate |
+| $3.65/day | World Bank | - | Working poverty threshold |
+| $6.85/day | World Bank | - | Upper middle-income threshold |
+
+Countries: 15 African nations
+Years: 2018-2025
+"""
+
+import numpy as np
+import pandas as pd
+from pathlib import Path
+
+
+COUNTRIES = [
+    "Nigeria", "Ethiopia", "Tanzania", "Kenya", "Uganda", 
+    "Mozambique", "Ghana", "Cote d'Ivoire", "Cameroon", "Senegal",
+    "Zambia", "Malawi", "Rwanda", "Burkina Faso", "Mali"
+]
+
+YEARS = list(range(2018, 2026))
+
+BASE_RATES = {
+    "below_2.15": 0.28,
+    "below_3.00": 0.46,
+    "below_3.65": 0.58,
+    "below_6.85": 0.82
+}
+
+COUNTRY_ADJUSTMENTS = {
+    "Nigeria": {"below_2.15": 1.3, "below_3.00": 1.2, "below_3.65": 1.15, "below_6.85": 1.1},
+    "Ethiopia": {"below_2.15": 0.9, "below_3.00": 0.85, "below_3.65": 0.9, "below_6.85": 0.95},
+    "Tanzania": {"below_2.15": 0.85, "below_3.00": 0.9, "below_3.65": 0.9, "below_6.85": 0.9},
+    "Kenya": {"below_2.15": 0.6, "below_3.00": 0.55, "below_3.65": 0.5, "below_6.85": 0.45},
+    "Uganda": {"below_2.15": 0.75, "below_3.00": 0.8, "below_3.65": 0.85, "below_6.85": 0.9},
+    "Mozambique": {"below_2.15": 1.1, "below_3.00": 1.05, "below_3.65": 1.0, "below_6.85": 1.0},
+    "Ghana": {"below_2.15": 0.5, "below_3.00": 0.45, "below_3.65": 0.4, "below_6.85": 0.35},
+    "Cote d'Ivoire": {"below_2.15": 0.7, "below_3.00": 0.65, "below_3.65": 0.6, "below_6.85": 0.55},
+    "Cameroon": {"below_2.15": 0.8, "below_3.00": 0.75, "below_3.65": 0.7, "below_6.85": 0.65},
+    "Senegal": {"below_2.15": 0.65, "below_3.00": 0.6, "below_3.65": 0.55, "below_6.85": 0.5},
+    "Zambia": {"below_2.15": 0.9, "below_3.00": 0.85, "below_3.65": 0.8, "below_6.85": 0.75},
+    "Malawi": {"below_2.15": 1.0, "below_3.00": 0.95, "below_3.65": 0.9, "below_6.85": 0.85},
+    "Rwanda": {"below_2.15": 0.7, "below_3.00": 0.65, "below_3.65": 0.6, "below_6.85": 0.55},
+    "Burkina Faso": {"below_2.15": 0.85, "below_3.00": 0.8, "below_3.65": 0.75, "below_6.85": 0.7},
+    "Mali": {"below_2.15": 0.95, "below_3.00": 0.9, "below_3.65": 0.85, "below_6.85": 0.8}
+}
+
+
+def dag_sample(probabilities, n_samples, rng):
+    """DAG-based sampling with probability weighting."""
+    indices = rng.choice(len(probabilities), size=n_samples, p=probabilities, replace=True)
+    unique, counts = np.unique(indices, return_counts=True)
+    return dict(zip(unique, counts))
+
+
+def generate_dataset(scenario: str, seed: int) -> pd.DataFrame:
+    """Generate poverty headcount dataset for Africa."""
+    np.random.seed(seed)
+    rng = np.random.default_rng(seed)
+    
+    scenario_config = {
+        "low_burden": {"n": 4000, "variance_scale": 0.8},
+        "moderate": {"n": 5000, "variance_scale": 1.0},
+        "high_burden": {"n": 6000, "variance_scale": 1.2}
+    }
+    
+    config = scenario_config[scenario]
+    n = config["n"]
+    scale = config["variance_scale"]
+    
+    records = []
+    
+    for year in YEARS:
+        year_factor = 1.0 - (year - 2018) * 0.015
+        
+        for country in COUNTRIES:
+            adj = COUNTRY_ADJUSTMENTS[country]
+            
+            for threshold in ["below_2.15", "below_3.00", "below_3.65", "below_6.85"]:
+                base_rate = BASE_RATES[threshold]
+                adjusted_rate = min(0.99, base_rate * adj[threshold] * year_factor * scale)
+                adjusted_rate = max(0.01, adjusted_rate)
+                
+                n_country_year = n // (len(YEARS) * len(COUNTRIES))
+                poverty_count = int(n_country_year * adjusted_rate)
+                
+                for _ in range(n_country_year):
+                    records.append({
+                        "country": country,
+                        "year": year,
+                        "poverty_line": threshold,
+                        "headcount_ratio": adjusted_rate + rng.normal(0, 0.02 * scale),
+                        "population_below": poverty_count * 1000,
+                        "total_population": n_country_year * 1000
+                    })
+    
+    df = pd.DataFrame(records)
+    df["headcount_ratio"] = df["headcount_ratio"].clip(0, 1)
+    
+    return df
+
+
+def main():
+    output_dir = Path(__file__).parent
+    output_dir.mkdir(parents=True, exist_ok=True)
+    
+    scenarios = {
+        "low_burden": 42,
+        "moderate": 43,
+        "high_burden": 44
+    }
+    
+    for scenario, seed in scenarios.items():
+        df = generate_dataset(scenario, seed)
+        df.to_csv(output_dir / f"poverty_headcount_{scenario}.csv", index=False)
+        print(f"Generated {scenario}: {len(df)} rows, seed={seed}")
+
+
+if __name__ == "__main__":
+    main()