electricsheepafrica/nigerian-banking-mobile-money

Nigerian Mobile Money Wallet Transactions

Dataset Type: Banking & Finance - Digital Payments
Version: 1.0
License: Apache 2.0
Language: English
Geography: Nigeria

---

Dataset Description

Synthetic Nigerian mobile money wallet transactions with fraud and churn labels. Covers major providers (OPay, PalmPay, MTN MoMo, Airtel Money) with realistic P2P, airtime, bill payment, and cash-in/out patterns.

Use Cases:

• Mobile money fraud detection
• Customer churn prediction
• Financial inclusion analysis
• Agent network optimization
• Digital wallet product development

---

Dataset Statistics

• Rows: 4,000,000 (pilot: 10,000)
• Columns: 13
• Time Range: 2024-01-01 to 2024-06-30 (6 months)
• Fraud Prevalence: 1.5%
• Churn Prevalence: 6%

---

Schema

Column	Type	Description
transaction_id	string	Unique transaction identifier (UUID)
wallet_id	string	Wallet identifier (WLT-########)
timestamp	datetime	Transaction timestamp (Africa/Lagos)
transaction_type	category	p2p_send, p2p_receive, cashin, cashout, airtime, data, billpay
amount_ngn	float	Transaction amount in Naira
fee_ngn	float	Transaction fee
balance_after_ngn	float	Wallet balance after transaction
agent_id	string	Agent ID (for cashin/cashout)
channel	category	ussd, app, agent, web
device_os	category	android, feature_phone, ios
kyc_tier	category	tier1, tier2, tier3
fraud_flag	bool	LABEL: True if fraudulent
churn_30d	bool	LABEL: True if churned in 30 days

---

Label Distribution

Fraud Flag

• Positive: 1.5%
• Negative: 98.5%

Fraud Drivers:

• High Cashout Ratio (cashout/cashin >2): 12x risk
• New Wallet (<7 days old): 5x risk
• High Velocity (>20 txns/24h): 15x risk
• Night Hours: 3x risk
• Round Amounts (₦50k, ₦100k): 4x risk

Churn (30 days)

• Positive: 6%
• Negative: 94%

Churn Drivers:

• Dormant (14+ days no activity): 8x risk
• Low Diversity (only airtime): 3x risk
• Failed Transactions (>3 failures): 4x risk
• Tier 1 Only (never upgraded): 2.5x risk
• No Bill Payments: 2x risk

---

Data Distributions

Transaction Types

Type	Percentage	Description
Airtime	25%	Mobile phone top-up
P2P Send	20%	Send money to another wallet
P2P Receive	20%	Receive money from another wallet
Bill Payment	15%	Utilities, cable TV, internet
Cash-in	10%	Deposit to wallet
Cash-out	8%	Withdraw from wallet
Data Bundle	2%	Internet data purchase

Channels

Channel	Percentage	Description
USSD	50%	*99# codes (no smartphone needed)
App	35%	Smartphone apps
Agent	13%	Through banking agents
Web	2%	Web portal

Device OS

OS	Percentage
Android	70%
Feature Phone	25%
iOS	5%

KYC Tiers

Tier	Daily Limit	Percentage
Tier 1	₦50,000	60%
Tier 2	₦200,000	35%
Tier 3	₦5,000,000	5%

---

Mobile Money Providers

Market Share

Provider	Share	Features
OPay	35%	Cashback rewards (1%)
PalmPay	30%	Referral bonuses (₦500)
MTN MoMo	20%	Network advantage (linked to MTN)
Airtel Money	10%	Airtel network integration
9PSB	5%	9mobile payment service

---

Nigerian Context

Usage Patterns

✅ Airtime Dominant: Small frequent airtime purchases (₦200-₦1,500)
✅ Family Transfers: 65% send to 3-5 family members regularly
✅ Agent Preference: Urban 75% cashout via agents, Rural 45%
✅ USSD Heavy: 50% still use USSD (*99#) over apps
✅ Bill Payments: DSTV, GOtv, EKEDC, IKEDC, internet

Financial Inclusion

• Unbanked Access: Mobile money reaches 30M+ unbanked Nigerians
• Rural Penetration: Feature phone support enables rural access
• Low Barriers: Tier 1 KYC requires only phone number
• Micro-transactions: Enables ₦50-₦500 transactions

Transaction Patterns

• Morning Peak: 9-12am (before work)
• Evening Peak: 6-8pm (after work, family transfers)
• Weekend Dip: -20% on Sundays
• Month-End Surge: +50% last week of month (salary days)

---

Files

mobile_money/
├── README.md
├── nigerian_mobile_money_pilot.parquet (10k, 0.62 MB)
├── nigerian_mobile_money.parquet (4M, ~320 MB) - Coming Soon
├── nigerian_mobile_money.csv (4M, ~640 MB) - Coming Soon
└── mobile_money_sample.csv (100 rows)

---

Usage Example

import pandas as pd
from sklearn.ensemble import RandomForestClassifier

# Load data
df = pd.read_parquet('nigerian_mobile_money.parquet')

# Fraud detection features
df['is_cashout'] = (df['transaction_type'] == 'cashout').astype(int)
df['is_night'] = ((df['timestamp'].dt.hour >= 22) | 
                  (df['timestamp'].dt.hour <= 6)).astype(int)
df['is_round'] = df['amount_ngn'].isin([50000, 100000, 200000]).astype(int)
df['hour'] = df['timestamp'].dt.hour

fraud_features = ['amount_ngn', 'fee_ngn', 'is_cashout', 
                   'is_night', 'is_round', 'hour']

X_fraud = df[fraud_features]
y_fraud = df['fraud_flag']

# Train fraud model
fraud_model = RandomForestClassifier(n_estimators=100)
fraud_model.fit(X_fraud, y_fraud)

# Churn prediction features
df['balance_low'] = (df['balance_after_ngn'] < 1000).astype(int)
df['is_tier1'] = (df['kyc_tier'] == 'tier1').astype(int)
df['only_airtime'] = (df['transaction_type'] == 'airtime').astype(int)

churn_features = ['balance_low', 'is_tier1', 'only_airtime', 'amount_ngn']

X_churn = df[churn_features]
y_churn = df['churn_30d']

# Train churn model
churn_model = RandomForestClassifier(n_estimators=100)
churn_model.fit(X_churn, y_churn)

---

Citation

@dataset{nigerian_mobile_money_2025,
  author = {Electric Sheep Africa},
  title = {Nigerian Mobile Money Wallet Transactions Dataset},
  year = {2025},
  publisher = {Hugging Face}
}

---

Last Updated: 2025-10-19
Status: ✅ Pilot Validated