transactional-data / data_simulator.py
Ymak7's picture
Upload 2 files
1f5d888 verified
import random
import json
import uuid
from datetime import datetime, timedelta
# Define laundering patterns
PATTERNS = ["Fan-Out", "Fan-In", "Cycle", "Bipartite", "Stack", "Scatter Gather", "Gather Scatter", "Random"]
TRANSACTION_TYPES = ["Credit", "Debit", "Wire", "Card", "Cash"]
TIME_PERIODS = ["Morning", "Afternoon", "Night"]
COUNTRIES = ["USA", "UK", "Canada", "Germany", "India", "China", "UAE", "Australia"]
# Store transaction history
transaction_history = {}
def create_transaction(aml_flag=None):
"""Generate a simulated transaction"""
sender = f"A{random.randint(1, 1000)}"
receiver = f"A{random.randint(1, 1000)}"
amount = random.randint(50, 10000)
transaction_id = str(uuid.uuid4())
# Random timestamp (past 6 months)
timestamp = (datetime.now() - timedelta(days=random.randint(0, 180))).isoformat()
currency = "USD"
transaction_type = random.choice(TRANSACTION_TYPES)
sender_ip = random.choice(COUNTRIES)
receiver_ip = random.choice(COUNTRIES)
# Risk Score & AML Flag (Ensuring 50% AML and 50% Normal Transactions)
if aml_flag is None:
risk_score = round(random.uniform(0, 1), 2)
aml_flag = 1 if risk_score > 0.7 else 0
else:
risk_score = round(random.uniform(0.8, 1), 2) if aml_flag == 1 else round(random.uniform(0, 0.3), 2)
# Track transaction history
sender_history = transaction_history.get(sender, [])
sender_history.append((amount, timestamp))
transaction_history[sender] = sender_history[-30:] # Store last 30 transactions
# Compute Behavioral & Network Features
last_24h_tx = [tx for tx in sender_history if datetime.fromisoformat(tx[1]) > datetime.now() - timedelta(days=1)]
last_7d_tx = [tx for tx in sender_history if datetime.fromisoformat(tx[1]) > datetime.now() - timedelta(days=7)]
last_30d_tx = sender_history
daily_tx_count = len(last_24h_tx)
weekly_tx_count = len(last_7d_tx)
monthly_tx_count = len(last_30d_tx)
total_tx_volume = sum(tx[0] for tx in last_30d_tx)
avg_tx_amount = total_tx_volume / len(last_30d_tx) if last_30d_tx else amount
tx_amount_deviation = abs(amount - avg_tx_amount)
new_account_flag = 1 if len(sender_history) < 3 else 0
dormant_to_active_flag = 1 if random.random() < 0.05 else 0 # 5% chance
# Transaction Metadata
weekend_flag = 1 if datetime.fromisoformat(timestamp).weekday() >= 5 else 0
repeated_amount_flag = 1 if sum(1 for tx in sender_history if tx[0] == amount) > 1 else 0
time_of_day = random.choice(TIME_PERIODS)
return {
"TransactionID": transaction_id,
"Timestamp": timestamp,
"SenderAccount": sender,
"ReceiverAccount": receiver,
"Amount": amount,
"Currency": currency,
"TransactionType": transaction_type,
"AML_Flag": aml_flag,
"RiskScore": risk_score,
"DailyTransactionCount": daily_tx_count,
"WeeklyTransactionCount": weekly_tx_count,
"MonthlyTransactionCount": monthly_tx_count,
"TotalTransactionVolume": total_tx_volume,
"AverageTransactionAmount": avg_tx_amount,
"TransactionAmountDeviation": tx_amount_deviation,
"NewAccountFlag": new_account_flag,
"DormantToActiveFlag": dormant_to_active_flag,
"SenderIPLocation": sender_ip,
"ReceiverIPLocation": receiver_ip,
"WeekendFlag": weekend_flag,
"RepeatedAmountFlag": repeated_amount_flag,
"TimeOfDay": time_of_day
}
# Generate 50% AML and 50% Normal Transactions
aml_transactions = [create_transaction(aml_flag=1) for _ in range(5000)] # 50% AML
normal_transactions = [create_transaction(aml_flag=0) for _ in range(5000)] # 50% Normal
transactions = aml_transactions + normal_transactions
random.shuffle(transactions)
# Save to JSON file
with open("simulated_transactions.json", "w") as f:
json.dump(transactions, f, indent=4)
print("✅ Balanced simulated dataset created: simulated_transactions.json")