generate_dataset.py

"""Generate synthetic online pharmacy & unregistered medicines dataset for SSA.

Research-based parameterization:
- WHO: 50% of medicines purchased online from illegal sites are
  falsified; growing e-pharmacy market in Africa.
- SAHPRA: Fake medicines common; painkillers, antibiotics, antimalarials,
  ARVs, sexual stimulants most counterfeited online.
- SSA context: Limited regulation of online pharmacies; social media
  sales; WhatsApp/Facebook marketplace; cross-border e-commerce.
"""

from __future__ import annotations

from pathlib import Path

import numpy as np
import pandas as pd

SEED = 42
N_PER_SCENARIO = 10_000

YEAR_RANGE = np.arange(2010, 2025)
YEAR_WEIGHTS = np.linspace(0.85, 1.3, len(YEAR_RANGE))
YEAR_WEIGHTS = YEAR_WEIGHTS / YEAR_WEIGHTS.sum()

SCENARIOS = {
    "social_media_marketplace": {
        "setting_probs": {"whatsapp": 0.30, "facebook_marketplace": 0.25,
                          "instagram": 0.20, "telegram": 0.15, "tiktok": 0.10},
        "product_probs": {"sexual_stimulant": 0.20, "weight_loss": 0.15,
                          "antibiotic": 0.15, "analgesic": 0.12,
                          "skin_lightening": 0.10, "supplement": 0.10,
                          "antimalarial": 0.08, "other": 0.10},
        "sf_prevalence": 0.45,
        "unregistered_pct": 0.70,
        "no_prescription_pct": 0.95,
        "pharmacist_involved_pct": 0.02,
        "verified_seller_pct": 0.05,
    },
    "dedicated_online_pharmacy": {
        "setting_probs": {"dedicated_website": 0.35, "mobile_app": 0.25,
                          "aggregator_platform": 0.20, "telemedicine_linked": 0.20},
        "product_probs": {"antibiotic": 0.15, "antihypertensive": 0.12,
                          "antidiabetic": 0.10, "analgesic": 0.12,
                          "contraceptive": 0.08, "ARV": 0.05,
                          "supplement": 0.15, "sexual_stimulant": 0.10,
                          "other": 0.13},
        "sf_prevalence": 0.20,
        "unregistered_pct": 0.40,
        "no_prescription_pct": 0.60,
        "pharmacist_involved_pct": 0.25,
        "verified_seller_pct": 0.30,
    },
    "cross_border_ecommerce": {
        "setting_probs": {"alibaba_aliexpress": 0.30, "jumia_marketplace": 0.20,
                          "international_website": 0.25, "dark_web": 0.10,
                          "unknown_platform": 0.15},
        "product_probs": {"antibiotic": 0.15, "sexual_stimulant": 0.15,
                          "controlled_substance": 0.10, "weight_loss": 0.10,
                          "supplement": 0.12, "antimalarial": 0.08,
                          "skin_lightening": 0.10, "other": 0.20},
        "sf_prevalence": 0.50,
        "unregistered_pct": 0.80,
        "no_prescription_pct": 0.90,
        "pharmacist_involved_pct": 0.01,
        "verified_seller_pct": 0.03,
    },
}

SCENARIO_FILES = {
    "social_media_marketplace": "online_social_media.csv",
    "dedicated_online_pharmacy": "online_dedicated.csv",
    "cross_border_ecommerce": "online_cross_border.csv",
}


def _choice(rng, prob_map):
    keys = list(prob_map.keys())
    weights = np.array(list(prob_map.values()), dtype=float)
    weights = weights / weights.sum()
    return rng.choice(keys, p=weights)


def _simulate_scenario(name, params, seed):
    rng = np.random.default_rng(seed)
    records = []

    for idx in range(N_PER_SCENARIO):
        year = int(rng.choice(YEAR_RANGE, p=YEAR_WEIGHTS))
        platform = _choice(rng, params["setting_probs"])
        product = _choice(rng, params["product_probs"])
        age = int(np.clip(rng.normal(28, 10), 15, 65))
        sex = rng.choice(["male", "female"], p=[0.50, 0.50])

        no_prescription = int(rng.random() < params["no_prescription_pct"])
        self_medication = int(no_prescription and rng.random() < 0.80)
        pharmacist_involved = int(rng.random() < params["pharmacist_involved_pct"])
        verified_seller = int(rng.random() < params["verified_seller_pct"])

        # Registration status
        unregistered = int(rng.random() < params["unregistered_pct"])
        nmra_approved = int(not unregistered and rng.random() < 0.60)
        manufacturer_known = int(rng.random() < 0.40)
        batch_traceable = int(manufacturer_known and rng.random() < 0.30)

        # Quality
        is_sf = int(rng.random() < params["sf_prevalence"])
        is_falsified = int(is_sf and rng.random() < 0.50)
        is_substandard = int(is_sf and not is_falsified)
        no_active_ingredient = int(is_falsified and rng.random() < 0.30)
        wrong_ingredient = int(is_falsified and rng.random() < 0.15)
        contaminated = int(is_sf and rng.random() < 0.10)

        # Delivery
        delivery_method = rng.choice(["courier", "postal", "pickup", "rider"],
                                      p=[0.30, 0.20, 0.20, 0.30])
        cold_chain_needed = int(product in ("vaccine", "ARV") and rng.random() < 0.30)
        cold_chain_maintained = int(cold_chain_needed and rng.random() < 0.10)
        delivery_days = int(np.clip(rng.exponential(5), 1, 30))

        # Health outcomes
        adr_occurred = int(is_sf and rng.random() < 0.08)
        treatment_failure = int(is_sf and rng.random() < 0.12)
        hospitalisation = int((adr_occurred or treatment_failure) and rng.random() < 0.10)
        delayed_care = int(self_medication and rng.random() < 0.15)

        # Regulation & enforcement
        reported_to_nmra = int(is_sf and rng.random() < 0.02)
        platform_removed = int(is_sf and rng.random() < 0.05)
        consumer_aware_risk = int(rng.random() < 0.10)
        checked_registration = int(rng.random() < 0.05)
        price_usd = float(np.clip(rng.lognormal(np.log(3), 0.8), 0.20, 100))

        any_adverse = int(adr_occurred or treatment_failure or delayed_care)

        record = {
            "record_id": f"{name[:3].upper()}-{idx:05d}",
            "scenario": name,
            "year": year,
            "platform": platform,
            "product": product,
            "age": age,
            "sex": sex,
            "no_prescription": no_prescription,
            "self_medication": self_medication,
            "pharmacist_involved": pharmacist_involved,
            "verified_seller": verified_seller,
            "unregistered": unregistered,
            "nmra_approved": nmra_approved,
            "manufacturer_known": manufacturer_known,
            "batch_traceable": batch_traceable,
            "is_substandard_falsified": is_sf,
            "is_falsified": is_falsified,
            "no_active_ingredient": no_active_ingredient,
            "wrong_ingredient": wrong_ingredient,
            "contaminated": contaminated,
            "delivery_method": delivery_method,
            "delivery_days": delivery_days,
            "adr_occurred": adr_occurred,
            "treatment_failure": treatment_failure,
            "hospitalisation": hospitalisation,
            "delayed_care": delayed_care,
            "reported_to_nmra": reported_to_nmra,
            "consumer_aware_risk": consumer_aware_risk,
            "price_usd": round(price_usd, 2),
            "any_adverse": any_adverse,
        }
        records.append(record)

    return pd.DataFrame(records)


def main():
    output_dir = Path("data")
    output_dir.mkdir(parents=True, exist_ok=True)
    for idx, (name, params) in enumerate(SCENARIOS.items()):
        df = _simulate_scenario(name, params, SEED + idx * 211)
        df.to_csv(output_dir / SCENARIO_FILES[name], index=False)
        print(f"Saved {name} -> {SCENARIO_FILES[name]}")


if __name__ == "__main__":
    main()