generate_dataset.py

"""Generate African Mobile Subscriber Data for 15 SSA countries across 3 scenarios."""

import csv
import json
import os
import random
from dataclasses import dataclass, field
from typing import Dict, List, Tuple

SEED = 42
random.seed(SEED)

COUNTRIES = [
    "Nigeria", "South Africa", "Kenya", "Ghana", "Tanzania",
    "Ethiopia", "Uganda", "Cote d'Ivoire", "Senegal", "Angola",
    "Mozambique", "Rwanda", "Cameroon", "Madagascar", "Zambia",
]

SCENARIOS = ["baseline", "5g_rollout", "market_saturation"]
RECORDS_PER_SCENARIO = 10000

YEARS = [2024, 2025, 2026]
QUARTERS = ["Q1", "Q2", "Q3", "Q4"]

TECHNOLOGIES = ["2g", "3g", "4g", "5g"]

# Country-specific operator profiles
OPERATOR_PROFILES: Dict[str, List[dict]] = {
    "Nigeria": [
        {"name": "MTN Nigeria", "base_subscribers": 85.0, "base_market_share": 38.0, "base_arpu": 4.8},
        {"name": "Glo", "base_subscribers": 60.0, "base_market_share": 27.0, "base_arpu": 3.9},
        {"name": "Airtel Africa (Nigeria)", "base_subscribers": 55.0, "base_market_share": 25.0, "base_arpu": 4.2},
        {"name": "9mobile", "base_subscribers": 20.0, "base_market_share": 10.0, "base_arpu": 3.5},
    ],
    "South Africa": [
        {"name": "MTN South Africa", "base_subscribers": 35.0, "base_market_share": 35.0, "base_arpu": 6.5},
        {"name": "Vodacom", "base_subscribers": 40.0, "base_market_share": 40.0, "base_arpu": 7.2},
        {"name": "Cell C", "base_subscribers": 12.0, "base_market_share": 12.0, "base_arpu": 4.8},
        {"name": "Telkom Mobile", "base_subscribers": 13.0, "base_market_share": 13.0, "base_arpu": 5.1},
    ],
    "Kenya": [
        {"name": "Safaricom", "base_subscribers": 42.0, "base_market_share": 63.0, "base_arpu": 5.5},
        {"name": "Airtel Kenya", "base_subscribers": 18.0, "base_market_share": 27.0, "base_arpu": 3.2},
        {"name": "Telkom Kenya", "base_subscribers": 7.0, "base_market_share": 10.0, "base_arpu": 2.8},
    ],
    "Ghana": [
        {"name": "MTN Ghana", "base_subscribers": 25.0, "base_market_share": 55.0, "base_arpu": 4.0},
        {"name": "Vodafone Ghana", "base_subscribers": 10.0, "base_market_share": 22.0, "base_arpu": 3.2},
        {"name": "AirtelTigo", "base_subscribers": 8.0, "base_market_share": 18.0, "base_arpu": 2.9},
    ],
    "Tanzania": [
        {"name": "Vodacom Tanzania", "base_subscribers": 16.0, "base_market_share": 32.0, "base_arpu": 3.5},
        {"name": "Airtel Tanzania", "base_subscribers": 14.0, "base_market_share": 28.0, "base_arpu": 3.0},
        {"name": "Tigo Tanzania", "base_subscribers": 12.0, "base_market_share": 24.0, "base_arpu": 3.1},
        {"name": "Halotel", "base_subscribers": 5.0, "base_market_share": 10.0, "base_arpu": 2.5},
        {"name": "TTCL", "base_subscribers": 3.0, "base_market_share": 6.0, "base_arpu": 2.2},
    ],
    "Ethiopia": [
        {"name": "Ethio Telecom", "base_subscribers": 70.0, "base_market_share": 85.0, "base_arpu": 2.8},
        {"name": "Safaricom Ethiopia", "base_subscribers": 12.0, "base_market_share": 15.0, "base_arpu": 3.5},
    ],
    "Uganda": [
        {"name": "MTN Uganda", "base_subscribers": 16.0, "base_market_share": 50.0, "base_arpu": 2.9},
        {"name": "Airtel Uganda", "base_subscribers": 12.0, "base_market_share": 37.0, "base_arpu": 2.6},
        {"name": "Lycamobile Uganda", "base_subscribers": 2.0, "base_market_share": 6.0, "base_arpu": 2.1},
    ],
    "Cote d'Ivoire": [
        {"name": "Orange Cote d'Ivoire", "base_subscribers": 16.0, "base_market_share": 47.0, "base_arpu": 3.4},
        {"name": "MTN Cote d'Ivoire", "base_subscribers": 12.0, "base_market_share": 35.0, "base_arpu": 3.1},
        {"name": "Moov Africa", "base_subscribers": 4.0, "base_market_share": 12.0, "base_arpu": 2.7},
    ],
    "Senegal": [
        {"name": "Orange Senegal", "base_subscribers": 12.0, "base_market_share": 50.0, "base_arpu": 3.6},
        {"name": "Free Senegal", "base_subscribers": 6.0, "base_market_share": 25.0, "base_arpu": 2.8},
        {"name": "Expresso", "base_subscribers": 3.0, "base_market_share": 13.0, "base_arpu": 2.3},
    ],
    "Angola": [
        {"name": "Unitel", "base_subscribers": 14.0, "base_market_share": 60.0, "base_arpu": 3.8},
        {"name": "Movicel", "base_subscribers": 6.0, "base_market_share": 26.0, "base_arpu": 3.0},
    ],
    "Mozambique": [
        {"name": "Vodacom Mozambique", "base_subscribers": 10.0, "base_market_share": 42.0, "base_arpu": 2.8},
        {"name": "Movitel", "base_subscribers": 8.0, "base_market_share": 33.0, "base_arpu": 2.5},
        {"name": "Tmcel", "base_subscribers": 4.0, "base_market_share": 17.0, "base_arpu": 2.1},
    ],
    "Rwanda": [
        {"name": "MTN Rwanda", "base_subscribers": 7.0, "base_market_share": 60.0, "base_arpu": 3.2},
        {"name": "Airtel Rwanda", "base_subscribers": 4.0, "base_market_share": 34.0, "base_arpu": 2.6},
    ],
    "Cameroon": [
        {"name": "MTN Cameroon", "base_subscribers": 12.0, "base_market_share": 45.0, "base_arpu": 3.3},
        {"name": "Orange Cameroon", "base_subscribers": 10.0, "base_market_share": 37.0, "base_arpu": 3.1},
        {"name": "Nexttel", "base_subscribers": 3.0, "base_market_share": 11.0, "base_arpu": 2.4},
    ],
    "Madagascar": [
        {"name": "Telma", "base_subscribers": 8.0, "base_market_share": 45.0, "base_arpu": 2.2},
        {"name": "Orange Madagascar", "base_subscribers": 6.0, "base_market_share": 34.0, "base_arpu": 2.5},
        {"name": "Airtel Madagascar", "base_subscribers": 3.0, "base_market_share": 17.0, "base_arpu": 2.0},
    ],
    "Zambia": [
        {"name": "MTN Zambia", "base_subscribers": 9.0, "base_market_share": 48.0, "base_arpu": 3.1},
        {"name": "Airtel Zambia", "base_subscribers": 7.0, "base_market_share": 37.0, "base_arpu": 2.8},
        {"name": "Zamtel", "base_subscribers": 2.0, "base_market_share": 11.0, "base_arpu": 2.3},
    ],
}

# Mobile money penetration factors by country (Safaricom M-Pesa dominance in Kenya)
MOBILE_MONEY_FACTORS: Dict[str, float] = {
    "Nigeria": 0.35, "South Africa": 0.08, "Kenya": 0.82,
    "Ghana": 0.55, "Tanzania": 0.60, "Ethiopia": 0.20,
    "Uganda": 0.65, "Cote d'Ivoire": 0.50, "Senegal": 0.45,
    "Angola": 0.10, "Mozambique": 0.35, "Rwanda": 0.55,
    "Cameroon": 0.30, "Madagascar": 0.25, "Zambia": 0.40,
}

# Technology mix baselines per country (2g, 3g, 4g, 5g shares)
TECH_MIX: Dict[str, Dict[str, List[float]]] = {}
for c in COUNTRIES:
    if c in ("South Africa", "Kenya", "Nigeria"):
        TECH_MIX[c] = {
            "baseline": [5, 25, 60, 10],
            "5g_rollout": [3, 15, 52, 30],
            "market_saturation": [4, 22, 62, 12],
        }
    elif c in ("Ghana", "Tanzania", "Rwanda", "Senegal"):
        TECH_MIX[c] = {
            "baseline": [10, 30, 55, 5],
            "5g_rollout": [6, 20, 50, 24],
            "market_saturation": [8, 28, 58, 6],
        }
    else:
        TECH_MIX[c] = {
            "baseline": [18, 35, 45, 2],
            "5g_rollout": [10, 25, 50, 15],
            "market_saturation": [15, 32, 48, 5],
        }


def _scenario_multiplier(scenario: str, base: float) -> float:
    if scenario == "baseline":
        return base
    elif scenario == "5g_rollout":
        return base * 1.08
    else:
        return base * 1.15


def _scenario_subscribers(scenario: str, subs: float) -> float:
    if scenario == "baseline":
        return subs
    elif scenario == "5g_rollout":
        return subs * 0.95
    else:
        return subs * 1.25


def _rand_tech(tech_mix: List[float]) -> str:
    return random.choices(TECHNOLOGIES, weights=tech_mix, k=1)[0]


def generate_records() -> List[dict]:
    records = []
    record_id = 1

    for scenario in SCENARIOS:
        for _ in range(RECORDS_PER_SCENARIO):
            country = random.choice(COUNTRIES)
            operators = OPERATOR_PROFILES[country]
            op = random.choice(operators)
            year = random.choice(YEARS)
            quarter = random.choice(QUARTERS)

            tech_mix = TECH_MIX[country][scenario]
            technology = random.choices(TECHNOLOGIES, weights=tech_mix, k=1)[0]

            subs_base = _scenario_subscribers(scenario, op["base_subscribers"])
            active_subscribers = round(
                subs_base * random.uniform(0.85, 1.15), 2
            )

            arpu_base = _scenario_multiplier(scenario, op["base_arpu"])
            arpu = round(arpu_base * random.uniform(0.88, 1.12), 2)

            churn = round(random.uniform(1.5, 6.5) if scenario != "market_saturation" else random.uniform(2.5, 8.0), 2)

            prepaid_share = round(random.uniform(70, 98) if country != "South Africa" else random.uniform(45, 75), 2)

            if technology in ("4g", "5g"):
                data_rev = round(random.uniform(45, 75) if scenario != "market_saturation" else random.uniform(55, 82), 2)
            else:
                data_rev = round(random.uniform(20, 45), 2)
            voice_rev = round(100 - data_rev, 2)

            mm_factor = MOBILE_MONEY_FACTORS[country]
            if country == "Kenya" and op["name"] == "Safaricom":
                mm_factor = 0.82
            mm_subs = round(active_subscribers * mm_factor * random.uniform(0.7, 1.1), 2)

            if technology == "5g":
                data_usage = round(random.uniform(12, 35), 2)
            elif technology == "4g":
                data_usage = round(random.uniform(4, 15), 2)
            elif technology == "3g":
                data_usage = round(random.uniform(1, 5), 2)
            else:
                data_usage = round(random.uniform(0.2, 1.5), 2)

            if scenario == "5g_rollout":
                data_usage *= random.uniform(1.2, 1.6)

            coverage = round(random.uniform(75, 99) if country in ("South Africa", "Kenya") else random.uniform(55, 92), 2)
            if scenario == "5g_rollout" and technology == "5g":
                coverage = round(random.uniform(30, 65), 2)

            market_share = round(op["base_market_share"] * random.uniform(0.9, 1.1), 2)

            if technology == "5g":
                spectrum_eff = round(random.uniform(3.5, 5.0), 2)
            elif technology == "4g":
                spectrum_eff = round(random.uniform(2.0, 3.5), 2)
            elif technology == "3g":
                spectrum_eff = round(random.uniform(1.0, 2.0), 2)
            else:
                spectrum_eff = round(random.uniform(0.3, 1.0), 2)

            satisfaction = round(random.uniform(3.0, 4.8) if country in ("South Africa", "Kenya", "Rwanda") else random.uniform(2.5, 4.2), 2)

            records.append({
                "record_id": record_id,
                "country": country,
                "year": year,
                "quarter": quarter,
                "operator": op["name"],
                "technology": technology,
                "active_subscribers_millions": active_subscribers,
                "arpu_usd": arpu,
                "monthly_churn_pct": churn,
                "prepaid_share_pct": prepaid_share,
                "data_revenue_share_pct": data_rev,
                "voice_revenue_share_pct": voice_rev,
                "mobile_money_subscribers_millions": mm_subs,
                "data_usage_gb_per_user": round(data_usage, 2),
                "network_coverage_pct": coverage,
                "market_share_pct": market_share,
                "spectrum_efficiency_index": spectrum_eff,
                "customer_satisfaction_score": satisfaction,
                "scenario": scenario,
            })
            record_id += 1

    random.shuffle(records)
    return records


def write_csv(records: List[dict], path: str) -> None:
    os.makedirs(os.path.dirname(path), exist_ok=True)
    fieldnames = list(records[0].keys())
    with open(path, "w", newline="", encoding="utf-8") as f:
        writer = csv.DictWriter(f, fieldnames=fieldnames)
        writer.writeheader()
        writer.writerows(records)


def write_jsonl(records: List[dict], path: str) -> None:
    with open(path, "w", encoding="utf-8") as f:
        for r in records:
            f.write(json.dumps(r, ensure_ascii=False) + "\n")


def main():
    print("Generating 30,000 records across 15 SSA countries and 3 scenarios...")
    records = generate_records()

    csv_path = "telecom/african-mobile-subscriber-data/data/african_mobile_subscribers.csv"
    jsonl_path = "telecom/african-mobile-subscriber-data/data/african_mobile_subscribers.jsonl"

    write_csv(records, csv_path)
    print(f"  CSV: {csv_path} ({len(records)} rows)")

    write_jsonl(records, jsonl_path)
    print(f"  JSONL: {jsonl_path} ({len(records)} rows)")

    # Summary
    countries_in_data = sorted(set(r["country"] for r in records))
    operators_in_data = sorted(set(r["operator"] for r in records))
    print(f"\nSummary:")
    print(f"  Countries: {len(countries_in_data)}")
    print(f"  Operators: {len(operators_in_data)}")
    print(f"  Scenarios: {SCENARIOS}")
    print(f"  Total records: {len(records)}")
    for scenario in SCENARIOS:
        count = sum(1 for r in records if r["scenario"] == scenario)
        print(f"    {scenario}: {count}")


if __name__ == "__main__":
    main()