"""
Mobile Money Adoption in Africa - Dataset Generator

DAG Structure:
    country → urban_rural, income_quintile
    year → mobile_money_penetration
    urban_rural → account_ownership
    gender → account_ownership
    age_group → account_ownership, digital_literacy
    income_quintile → account_ownership, transaction_frequency
    account_ownership → active_use, transaction_types
    active_use → avg_transaction_value

Parameter Evidence Table:
| Parameter | Value | Source |
|-----------|-------|--------|
| SSA mobile money accounts | 1.1 billion | GSMA SOTIR 2025 |
| Global registered accounts | 2 billion | GSMA SOTIR 2025 |
| Active users globally | 500M+ | GSMA SOTIR 2025 |
| SSA GDP contribution | $190 billion (2023) | GSMA SOTIR 2025 |
| Kenya M-Pesa adoption | 84% | World Bank Findex 2025 |
| Uganda mobile money | 56% | World Bank Findex 2025 |
| Tanzania mobile money | 48% | World Bank Findex 2025 |
| Ghana mobile money | 41% | World Bank Findex 2025 |
| Nigeria mobile money | 12% | World Bank Findex 2025 |
| Urban adoption advantage | 1.5-2x rural | GSMA SOTIR 2025 |
| Gender gap | 8-15% lower for women | World Bank Findex 2025 |
| Active account rate | 35-45% of registered | GSMA SOTIR 2025 |
| Avg transaction value | $50-200 | GSMA SOTIR 2025 |
"""

import numpy as np
import pandas as pd
from pathlib import Path

COUNTRIES = [
    "Kenya", "Uganda", "Nigeria", "Ghana", "Tanzania", "Ethiopia",
    "Malawi", "Zambia", "Senegal", "Rwanda", "Niger", "Mali",
    "DRC", "Mozambique", "South Africa"
]

YEARS = list(range(2018, 2026))

COUNTRY_BASE_RATES = {
    "Kenya": 0.84, "Uganda": 0.56, "Tanzania": 0.48, "Ghana": 0.41,
    "Nigeria": 0.12, "Ethiopia": 0.08, "Malawi": 0.25, "Zambia": 0.32,
    "Senegal": 0.45, "Rwanda": 0.38, "Niger": 0.18, "Mali": 0.22,
    "DRC": 0.15, "Mozambique": 0.20, "South Africa": 0.28
}

PROVIDERS = {
    "Kenya": ["M-Pesa", "Airtel Money"],
    "Uganda": ["MTN Mobile Money", "Airtel Money"],
    "Tanzania": ["M-Pesa", "Tigo Pesa", "Airtel Money"],
    "Ghana": ["MTN Mobile Money", "Vodafone Cash", "AirtelTigo Money"],
    "Nigeria": ["Paga", "OPay", " PalmPay"],
    "Ethiopia": ["telebirr", "M-Birr"],
    "Malawi": ["Airtel Money", "TNM Mpamba"],
    "Zambia": ["MTN Mobile Money", "Airtel Money", "Zamtel Kwacha"],
    "Senegal": ["Orange Money", "Wave", "Wari"],
    "Rwanda": ["MTN Mobile Money", "Airtel Money"],
    "Niger": ["Airtel Money", "Moov Money"],
    "Mali": ["Orange Money", "Moov Money"],
    "DRC": ["M-Pesa", "Orange Money", "Airtel Money"],
    "Mozambique": ["M-Pesa", "e-Mola"],
    "South Africa": ["Vodapay", "MTN Mobile Money", "FNB eWallet"]
}

TRANSACTION_TYPES = [
    "airtime_purchase", "bill_payment", "person_to_person",
    "merchant_payment", "cash_withdrawal", "cash_deposit",
    "savings", "loan_repayment", "international_remittance"
]


def generate_scenario(n_samples: int, seed: int, scenario: str) -> pd.DataFrame:
    rng = np.random.default_rng(seed)
    
    country = rng.choice(COUNTRIES, n_samples)
    year = rng.choice(YEARS, n_samples)
    
    urban_rural = np.where(
        rng.random(n_samples) < 0.42,
        "urban", "rural"
    )
    
    gender = rng.choice(["male", "female"], n_samples, p=[0.48, 0.52])
    
    age = rng.integers(15, 75, n_samples)
    age_group = pd.cut(age, bins=[14, 25, 35, 45, 55, 75], 
                       labels=["15-24", "25-34", "35-44", "45-54", "55+"])
    
    income_quintile = rng.choice(["lowest", "second", "middle", "fourth", "highest"], 
                                  n_samples, p=[0.25, 0.22, 0.20, 0.18, 0.15])
    
    base_rates = np.array([COUNTRY_BASE_RATES[c] for c in country])
    
    year_adj = (year - 2018) * 0.025
    urban_adj = np.where(urban_rural == "urban", 0.15, 0.0)
    gender_adj = np.where(gender == "female", -0.08, 0.0)
    
    age_adj = np.where(
        (age >= 25) & (age <= 44), 0.10,
        np.where(age < 25, 0.05, -0.05)
    )
    
    quintile_adj = np.select(
        [income_quintile == "lowest", income_quintile == "second",
         income_quintile == "middle", income_quintile == "fourth",
         income_quintile == "highest"],
        [-0.15, -0.08, 0.0, 0.08, 0.15]
    )
    
    ownership_prob = np.clip(base_rates + year_adj + urban_adj + gender_adj + age_adj + quintile_adj, 0.05, 0.95)
    account_ownership = (rng.random(n_samples) < ownership_prob).astype(int)
    
    active_use = np.zeros(n_samples, dtype=int)
    active_mask = account_ownership == 1
    active_prob = np.where(urban_rural[active_mask] == "urban", 0.50, 0.38)
    active_use[active_mask] = (rng.random(active_mask.sum()) < active_prob).astype(int)
    
    n_transactions_monthly = np.zeros(n_samples)
    trans_mask = active_use == 1
    base_trans = np.where(urban_rural[trans_mask] == "urban", 12, 8)
    n_transactions_monthly[trans_mask] = rng.poisson(base_trans)
    
    avg_transaction_value = np.zeros(n_samples)
    avg_transaction_value[trans_mask] = rng.lognormal(
        mean=np.log(75) + np.where(urban_rural[trans_mask] == "urban", 0.3, 0),
        sigma=0.6,
        size=trans_mask.sum()
    )
    
    providers = np.empty(n_samples, dtype=object)
    prov_mask = account_ownership == 1
    for i, c in enumerate(country):
        if prov_mask[i]:
            providers[i] = rng.choice(PROVIDERS[c])
        else:
            providers[i] = "none"
    
    transaction_types_list = []
    for i in range(n_samples):
        if active_use[i] == 1:
            n_types = rng.integers(2, 6)
            types = rng.choice(TRANSACTION_TYPES, n_types, replace=False)
            transaction_types_list.append(",".join(types))
        else:
            transaction_types_list.append("none")
    
    digital_literacy = rng.integers(1, 11, n_samples)
    digital_literacy = np.clip(
        digital_literacy + 
        np.where(urban_rural == "urban", 1, 0) +
        np.where(age < 35, 1, 0),
        1, 10
    )
    
    years_active = np.zeros(n_samples, dtype=int)
    years_active[account_ownership == 1] = rng.integers(
        1, np.maximum(2, year[account_ownership == 1] - 2007),
        size=active_mask.sum()
    )
    
    df = pd.DataFrame({
        "country": country,
        "year": year,
        "urban_rural": urban_rural,
        "gender": gender,
        "age": age,
        "age_group": age_group,
        "income_quintile": income_quintile,
        "account_ownership": account_ownership,
        "active_use": active_use,
        "provider": providers,
        "n_transactions_monthly": n_transactions_monthly.astype(int),
        "avg_transaction_value_usd": np.round(avg_transaction_value, 2),
        "transaction_types": transaction_types_list,
        "digital_literacy_score": digital_literacy,
        "years_active": years_active,
        "scenario": scenario
    })
    
    return df


def main():
    output_dir = Path(__file__).parent
    
    scenarios = [
        ("low_burden", 4000, 42),
        ("moderate_burden", 5000, 43),
        ("high_burden", 6000, 44)
    ]
    
    all_dfs = []
    for scenario_name, n, seed in scenarios:
        df = generate_scenario(n, seed, scenario_name)
        all_dfs.append(df)
        
        output_path = output_dir / f"mobile_money_{scenario_name}.csv"
        df.to_csv(output_path, index=False)
        print(f"Generated {output_path}: {len(df)} records")
    
    combined = pd.concat(all_dfs, ignore_index=True)
    combined_path = output_dir / "mobile_money_combined.csv"
    combined.to_csv(combined_path, index=False)
    print(f"Generated {combined_path}: {len(combined)} records")


if __name__ == "__main__":
    main()