Abdulmajeedyahya/1M-Fraud-Labeled-Synthetic-Bank-Transactions

license: other tags: - tabular - finance - fraud-detection - synthetic-data - banking

1M Synthetic Fraud-Labeled Bank Transactions – Production-Ready Dataset

Full Dataset → 📥Download the Complete 1M Labeled Dataset Here

High-fidelity synthetic dataset simulating real-world banking transactions with realistic fraud patterns. Designed specifically for building robust, production-grade fraud detection models without any privacy or regulatory risks.

Dataset Overview

• Total Rows (Full Version): 1,000,000 transactions
• Fraud Label: Binary (isFraud — 0 = legitimate, 1 = fraudulent)
• Fully Synthetic: Generated to mirror the statistical properties and behavioral patterns of real American banking transactions (card payments, transfers, withdrawals, etc.)
• 100% PII-Free: No real customer data, account numbers, or personally identifiable information — fully compliant and safe for commercial use and model training

Why This Dataset Stands Out

• Realistic transaction features including amount, type, timestamps, balance changes, and merchant information
• Carefully engineered fraud scenarios that reflect common real-world attack vectors (account takeover, money mule, unusual spending patterns, etc.)
• Suitable for advanced experiments: imbalanced classification, anomaly detection, ensemble methods, deep learning, and explainable AI (XAI)
• Perfect for benchmarking models before deploying them in live financial environments

Files Included

• synthetic_sample.csv → Large preview sample (instant download) – Use this to explore the data and build initial models
• synthetic_transactions_1M.csv → Full dataset (1 million rows) – Available for purchase on Gumroad

Full Dataset → Download the Complete 1M Labeled Dataset Here

Ideal For

• Fraud Detection Systems
• Financial Machine Learning Research
• Building production-ready anti-fraud models (XGBoost, LightGBM, CatBoost, Neural Networks, etc.)
• Academic research and Kaggle competitions-style benchmarking

This dataset gives you the scale and realism needed to train models that actually perform well in production — without the headaches of real banking data access or compliance issues.

Keywords: fraud detection, synthetic data, bank transactions, machine learning dataset, imbalanced classification, tabular data, financial fraud, anomaly detection, production ML

---

Column	Description
transaction_id	Unique identifier for each transaction.
customer_id	Anonymized unique customer identifier.
customer_segment	Customer segmentation category (e.g., SMALL_BUSINESS, STUDENT, HIGH_NET_WORTH, MASS_RETAIL, FREQUENT_TRAVELER).
account_type	Type of bank account (CREDIT, CHECKING).
transaction_timestamp	Exact timestamp of the transaction (YYYY-MM-DD HH:MM:SS).
transaction_date	Date of the transaction (YYYY-MM-DD).
transaction_hour	Hour of the day (0-23) when the transaction occurred.
transaction_dow	Day of week (0 = Monday, 6 = Sunday).
is_weekend	Binary flag indicating whether the transaction occurred on a weekend (1 = weekend).
is_holiday	Binary flag indicating whether the transaction occurred on a public holiday (1 = holiday).
month	Month of the transaction (1-12).
quarter	Calendar quarter of the transaction (1-4).
amount	Transaction amount in the original currency.
currency	Currency code of the transaction (USD, EUR, JPY, CAD, GBP, SGD, etc.).
old_balance	Account balance before the transaction.
new_balance	Account balance after the transaction.
balance_delta_check	Difference between old and new balance (for validation).
merchant_name	Name of the merchant or recipient.
merchant_id	Unique merchant identifier.
mcc_code	Merchant Category Code (MCC) according to banking standards.
mcc_category	High-level category of the merchant (e.g., GROCERY, FUEL, ONLINE_RETAIL).
merchant_sector	Broader merchant sector (e.g., Retail, Financial Services, Travel).
payment_rail	Payment method / rail used (CREDIT_CARD, DEBIT_CARD, ACH_TRANSFER, SWIFT_WIRE, P2P_TRANSFER, etc.).
transaction_status	Final status of the transaction (APPROVED, DECLINED, PENDING).
is_international	Binary flag indicating whether the transaction is cross-border (1 = international).
is_round_amount	Binary flag indicating if the amount is a round number (psychological indicator).
device_id	Anonymized device identifier.
device_type	Type of device used (MOBILE_ANDROID, MOBILE_IOS, DESKTOP_WIN, POS_TERMINAL, ATM, etc.).
device_os	Operating system of the device.
trusted_device_count	Number of trusted devices associated with the customer.
new_device_flag	Binary flag indicating if the device is new/unknown for this customer.
ip_address	Anonymized IP address of the transaction.
city	City where the transaction originated.
country	Country code (ISO) of the transaction origin.
distance_from_home_km	Distance in kilometers from the customer’s usual location.
credit_limit	Customer’s credit limit (if applicable).
credit_utilization	Credit utilization ratio (0–1).
credit_score	Synthetic credit score of the customer.
account_age_days	Age of the customer’s account in days.
transaction_velocity_1h	Number of transactions made by the customer in the last hour.
chargeback_flag	Binary flag indicating whether a chargeback was filed later.
fraud_label	Target variable: Binary fraud label (0 = Legitimate, 1 = Fraud).
fraud_type	Detailed fraud category when fraud is detected (LEGITIMATE, CARD_TESTING, IMPOSSIBLE_TRAVEL, SMURFING, etc.).

---

License:
Preview sample is released under CC BY 4.0.
The full dataset is available under a commercial license via Gumroad purchase.