generate_dataset.py

"""
Livestock Health and Productivity - Sub-Saharan Africa
========================================================
Based on research:
- FAO 2024: Cattle mortality 8-15% in smallholder systems
- ILRI 2023: Milk yields 2-5L/day vs 15-25L potential
- World Bank 2023: Poultry mortality 15-30% in traditional systems
- AU-IBAR 2023: Disease prevalence 20-40% for endemic diseases
- GALVmed 2023: Vaccination coverage 10-30% for priority diseases

PARAMETER EVIDENCE TABLE
─────────────────────────────────────────────────────────────────────────
Parameter              │ Value Used       │ Source                    │ Year
───────────────────────┼──────────────────┼───────────────────────────┼──────
Cattle mortality       │ 8-15%            │ FAO 2024                  │ 2024
Milk yields            │ 2-5L/day         │ ILRI 2023                 │ 2023
Poultry mortality      │ 15-30%           │ World Bank 2023           │ 2023
Disease prevalence     │ 20-40%           │ AU-IBAR 2023              │ 2023
Vaccination coverage   │ 10-30%           │ GALVmed 2023              │ 2023
Goat mortality         │ 10-20%           │ FAO 2024                  │ 2024

Author: Electric Sheep Africa
"""

import numpy as np
import pandas as pd
import argparse
import os

np.random.default_rng(42)

COUNTRIES = ['Kenya', 'Uganda', 'Nigeria', 'Ghana', 'Tanzania', 'Ethiopia', 'Malawi', 'Zambia', 'Mali', 'Burkina Faso']
SPECIES = ['cattle', 'goats', 'sheep', 'poultry', 'pigs', 'donkeys']
PRODUCTION_SYSTEMS = ['extensive', 'semi_intensive', 'intensive', 'pastoral', 'agropastoral']
BREED_TYPES = ['local', 'crossbreed', 'exotic']
FEED_SOURCES = ['grazing', 'crop_residues', 'concentrates', 'fodder_crops', 'mixed']
HEALTH_ISSUES = ['tick_borne', 'respiratory', 'digestive', 'parasitic', 'metabolic', 'reproductive']
YEARS = list(range(2018, 2026))

SPECIES_BASE_MORTALITY = {
    'cattle': 0.12, 'goats': 0.15, 'sheep': 0.14, 'poultry': 0.22, 'pigs': 0.18, 'donkeys': 0.08
}

SPECIES_BASE_PRODUCTIVITY = {
    'cattle': {'milk_l_day': 3.5, 'weight_gain_kg_day': 0.4, 'calving_rate': 0.55},
    'goats': {'milk_l_day': 0.8, 'weight_gain_kg_day': 0.06, 'kidding_rate': 1.2},
    'sheep': {'milk_l_day': 0.5, 'weight_gain_kg_day': 0.08, 'lambing_rate': 1.1},
    'poultry': {'eggs_year': 80, 'weight_gain_kg_week': 0.15},
    'pigs': {'litter_size': 8, 'weight_gain_kg_day': 0.35},
    'donkeys': {'work_hours_day': 4, 'weight_gain_kg_day': 0.1}
}

DISEASE_PREVALENCE = {
    'tick_borne': 0.23, 'respiratory': 0.17, 'digestive': 0.14,
    'parasitic': 0.28, 'metabolic': 0.07, 'reproductive': 0.11
}

def sc(p, rng):
    a = np.array(list(p.values()))
    return rng.choice(list(p.keys()), p=a/a.sum())

def gen(n=5000, seed=42):
    rng = np.random.default_rng(seed)
    recs = []
    
    for i in range(n):
        country = rng.choice(COUNTRIES)
        year = rng.choice(YEARS)
        
        record_id = f"LVS-{country[:3].upper()}-{year}-{i+1:05d}"
        
        farm_size = rng.lognormal(0.2, 0.6)
        farm_size = np.clip(farm_size, 0.2, 30.0)
        
        production_system = sc({
            'extensive': 0.35, 'semi_intensive': 0.25, 'intensive': 0.10,
            'pastoral': 0.15, 'agropastoral': 0.15
        }, rng)
        
        primary_species = rng.choice(SPECIES, p=[0.30, 0.25, 0.15, 0.20, 0.07, 0.03])
        
        herd_size_base = {'cattle': 8, 'goats': 15, 'sheep': 12, 'poultry': 35, 'pigs': 6, 'donkeys': 2}[primary_species]
        herd_size = int(rng.lognormal(np.log(herd_size_base), 0.8))
        herd_size = np.clip(herd_size, 1, {'cattle': 100, 'goats': 200, 'sheep': 150, 'poultry': 500, 'pigs': 50, 'donkeys': 10}[primary_species])
        
        breed_type = sc({
            'local': 0.60, 'crossbreed': 0.30, 'exotic': 0.10
        }, rng)
        
        feed_source = sc({
            'grazing': 0.45, 'crop_residues': 0.20, 'concentrates': 0.10,
            'fodder_crops': 0.10, 'mixed': 0.15
        }, rng)
        
        if production_system in ['pastoral', 'extensive']:
            feed_source = 'grazing'
        elif production_system == 'intensive':
            feed_source = rng.choice(['concentrates', 'mixed'], p=[0.5, 0.5])
        
        water_access = rng.choice(['adequate', 'seasonal', 'limited', 'poor'], p=[0.35, 0.30, 0.25, 0.10])
        
        housing_type = rng.choice(['none', 'simple_shelter', 'improved_shelter', 'modern'], p=[0.35, 0.35, 0.20, 0.10])
        
        base_mortality = SPECIES_BASE_MORTALITY[primary_species]
        mortality_modifier = 1.0
        mortality_modifier *= 0.7 if production_system == 'intensive' else 1.2 if production_system in ['pastoral', 'extensive'] else 1.0
        mortality_modifier *= 0.8 if breed_type == 'local' else 1.1 if breed_type == 'exotic' else 1.0
        mortality_modifier *= 0.9 if housing_type in ['improved_shelter', 'modern'] else 1.1 if housing_type == 'none' else 1.0
        
        mortality_rate = base_mortality * mortality_modifier * rng.uniform(0.8, 1.2)
        mortality_rate = np.clip(mortality_rate, 0.03, 0.45)
        
        animals_died = int(herd_size * mortality_rate)
        
        disease_outbreak = rng.random() < 0.20
        
        primary_disease = None
        if disease_outbreak:
            primary_disease = rng.choice(list(DISEASE_PREVALENCE.keys()), 
                                        p=list(DISEASE_PREVALENCE.values()))
        
        disease_severity = rng.choice(['mild', 'moderate', 'severe'], p=[0.40, 0.40, 0.20]) if disease_outbreak else 'none'
        
        vaccination_coverage = rng.lognormal(np.log(0.20), 0.8)
        vaccination_coverage = np.clip(vaccination_coverage, 0.02, 0.60)
        
        animals_vaccinated = int(herd_size * vaccination_coverage)
        
        deworming_frequency = rng.choice(['never', 'yearly', 'biannual', 'quarterly'], p=[0.30, 0.35, 0.25, 0.10])
        
        tick_control = rng.random() < 0.45
        
        veterinary_access = rng.random() < 0.35
        
        vet_visits_year = rng.integers(0, 6) if veterinary_access else 0
        
        health_insurance = rng.random() < 0.05
        
        base_prod = SPECIES_BASE_PRODUCTIVITY[primary_species]
        
        prod_modifier = 1.0
        prod_modifier *= 1.3 if breed_type == 'exotic' else 1.1 if breed_type == 'crossbreed' else 1.0
        prod_modifier *= 0.85 if disease_outbreak else 1.0
        prod_modifier *= 0.9 if water_access in ['limited', 'poor'] else 1.0
        prod_modifier *= 1.15 if production_system == 'intensive' else 0.9 if production_system == 'pastoral' else 1.0
        prod_modifier *= 1.1 if feed_source in ['concentrates', 'mixed'] else 0.95 if feed_source == 'grazing' else 1.0
        
        if primary_species == 'cattle':
            milk_yield_l_day = base_prod['milk_l_day'] * prod_modifier * rng.uniform(0.8, 1.2)
            milk_yield_l_day = np.clip(milk_yield_l_day, 0.5, 20)
            lactation_length_days = int(rng.normal(250, 40))
            annual_milk_l = milk_yield_l_day * lactation_length_days
            weight_gain_kg_day = base_prod['weight_gain_kg_day'] * prod_modifier * rng.uniform(0.7, 1.3)
            calving_rate = base_prod['calving_rate'] * prod_modifier * rng.uniform(0.8, 1.1)
            calving_rate = np.clip(calving_rate, 0.3, 0.9)
            live_weight_kg = rng.normal(320, 60) if breed_type == 'local' else rng.normal(400, 70)
        elif primary_species == 'goats':
            milk_yield_l_day = base_prod['milk_l_day'] * prod_modifier * rng.uniform(0.7, 1.3)
            weight_gain_kg_day = base_prod['weight_gain_kg_day'] * prod_modifier
            kidding_rate = base_prod['kidding_rate'] * prod_modifier
            annual_milk_l = milk_yield_l_day * 180
            calving_rate = kidding_rate
            lactation_length_days = 180
            live_weight_kg = rng.normal(28, 8)
        elif primary_species == 'sheep':
            milk_yield_l_day = base_prod['milk_l_day']
            weight_gain_kg_day = base_prod['weight_gain_kg_day'] * prod_modifier
            annual_milk_l = 0
            lambing_rate = base_prod['lambing_rate']
            calving_rate = lambing_rate
            lactation_length_days = 0
            live_weight_kg = rng.normal(32, 7)
        elif primary_species == 'poultry':
            eggs_per_year = int(base_prod['eggs_year'] * prod_modifier * rng.uniform(0.7, 1.3))
            weight_gain_kg_week = base_prod['weight_gain_kg_week'] * prod_modifier
            weight_gain_kg_day = weight_gain_kg_week / 7
            milk_yield_l_day = 0
            annual_milk_l = 0
            calving_rate = 0
            lactation_length_days = 0
            live_weight_kg = rng.normal(1.8, 0.5)
        elif primary_species == 'pigs':
            litter_size = int(base_prod['litter_size'] * prod_modifier * rng.uniform(0.8, 1.2))
            weight_gain_kg_day = base_prod['weight_gain_kg_day'] * prod_modifier
            litters_per_year = rng.uniform(1.8, 2.3)
            milk_yield_l_day = 0
            annual_milk_l = 0
            calving_rate = litter_size
            lactation_length_days = 0
            live_weight_kg = rng.normal(80, 20)
        else:
            milk_yield_l_day = 0
            annual_milk_l = 0
            calving_rate = 0
            lactation_length_days = 0
            weight_gain_kg_day = base_prod['weight_gain_kg_day']
            live_weight_kg = rng.normal(140, 25)
        
        reproductive_health = rng.random() < 0.75
        fertility_issues = not reproductive_health and rng.random() < 0.25
        
        body_condition_score = rng.integers(1, 6)
        
        nutrition_status = 'adequate' if body_condition_score >= 3 else 'poor'
        
        supplementation = rng.random() < 0.30
        
        mineral_supplement = rng.random() < 0.20
        
        forage_quality = rng.choice(['poor', 'moderate', 'good'], p=[0.30, 0.45, 0.25])
        
        breeding_method = rng.choice(['natural', 'artificial_insemination', 'both'], p=[0.75, 0.10, 0.15])
        
        record_keeping = rng.random() < 0.20
        
        market_access = rng.choice(['farm_gate', 'local_market', 'regional', 'export'], p=[0.40, 0.35, 0.20, 0.05])
        
        price_per_head = live_weight_kg * rng.lognormal(np.log(2.5), 0.3)
        
        annual_revenue = herd_size * price_per_head * 0.3
        
        feed_cost = herd_size * rng.uniform(20, 80) if feed_source in ['concentrates', 'mixed'] else herd_size * rng.uniform(5, 20)
        health_cost = animals_vaccinated * 2 + vet_visits_year * 25
        labor_cost = herd_size * rng.uniform(5, 15)
        
        total_cost = feed_cost + health_cost + labor_cost
        
        net_income = annual_revenue - total_cost
        
        productivity_index = 50 + (body_condition_score - 2) * 10 + (10 if reproductive_health else -5)
        productivity_index = np.clip(productivity_index, 10, 100)
        
        health_index = 100 - mortality_rate * 200 - (15 if disease_outbreak else 0) + vaccination_coverage * 50
        health_index = np.clip(health_index, 10, 100)
        
        recs.append({
            'record_id': i + 1,
            'livestock_id': record_id,
            'country': country,
            'year': year,
            'farm_size_ha': round(farm_size, 2),
            'production_system': production_system,
            'primary_species': primary_species,
            'herd_size': herd_size,
            'breed_type': breed_type,
            'feed_source': feed_source,
            'water_access': water_access,
            'housing_type': housing_type,
            'mortality_rate_pct': round(mortality_rate * 100, 1),
            'animals_died': animals_died,
            'disease_outbreak': disease_outbreak,
            'primary_disease': primary_disease if primary_disease else 'none',
            'disease_severity': disease_severity,
            'vaccination_coverage_pct': round(vaccination_coverage * 100, 1),
            'animals_vaccinated': animals_vaccinated,
            'deworming_frequency': deworming_frequency,
            'tick_control': tick_control,
            'veterinary_access': veterinary_access,
            'vet_visits_per_year': vet_visits_year,
            'health_insurance': health_insurance,
            'milk_yield_l_day': round(milk_yield_l_day, 2) if milk_yield_l_day > 0 else 0,
            'lactation_length_days': lactation_length_days if lactation_length_days > 0 else 0,
            'annual_milk_l': round(annual_milk_l, 0) if annual_milk_l > 0 else 0,
            'weight_gain_kg_day': round(weight_gain_kg_day, 3) if primary_species in ['cattle', 'goats', 'sheep', 'pigs'] else 0,
            'live_weight_kg': round(live_weight_kg, 1),
            'reproductive_rate': round(calving_rate, 2) if calving_rate > 0 else 0,
            'reproductive_health': reproductive_health,
            'fertility_issues': fertility_issues,
            'body_condition_score': body_condition_score,
            'nutrition_status': nutrition_status,
            'supplementation': supplementation,
            'mineral_supplement': mineral_supplement,
            'forage_quality': forage_quality,
            'breeding_method': breeding_method,
            'record_keeping': record_keeping,
            'market_access': market_access,
            'price_per_head_usd': round(price_per_head, 2),
            'annual_revenue_usd': round(annual_revenue, 2),
            'feed_cost_usd': round(feed_cost, 2),
            'health_cost_usd': round(health_cost, 2),
            'labor_cost_usd': round(labor_cost, 2),
            'total_cost_usd': round(total_cost, 2),
            'net_income_usd': round(net_income, 2),
            'productivity_index': round(productivity_index, 1),
            'health_index': round(health_index, 1),
            'mortality_category': 'low' if mortality_rate < 0.10 else 'moderate' if mortality_rate < 0.20 else 'high',
            'intervention_priority': 'high' if mortality_rate > 0.20 or disease_outbreak else 'moderate' if mortality_rate > 0.12 else 'low'
        })
    
    return pd.DataFrame(recs)

if __name__ == "__main__":
    p = argparse.ArgumentParser()
    p.add_argument('--n', type=int, default=5000)
    p.add_argument('--output', type=str, default='.')
    a = p.parse_args()
    
    for sn, m, s in [('low_burden', 0.8, 42), ('moderate_burden', 1.0, 43), ('high_burden', 1.2, 44)]:
        d = gen(int(a.n * m), s)
        d['scenario'] = sn
        d.to_csv(os.path.join(a.output, f'livestock_health_productivity_africa_{sn}.csv'), index=False)
        print(f"Saved: livestock_health_productivity_africa_{sn}.csv, n={len(d)}")