""" generate_transactions.py Lecture 10 — End-to-End Capstone: Banking Fraud Detection Streaming Data Analytics | MBA Course Generates a 2000-row synthetic banking transaction dataset with ~5% deliberately planted fraud patterns for classroom use. Usage: python generate_transactions.py Output: transactions.csv """ import csv import random import math from datetime import datetime, timedelta random.seed(42) # --------------------------------------------------------------------------- # 50 Synthetic Users # --------------------------------------------------------------------------- USERS = [ ("USR-001","Aarav Sharma","4521"), ("USR-002","Priya Mehta","7834"), ("USR-003","Rohan Gupta","2290"), ("USR-004","Anjali Singh","9012"), ("USR-005","Vikram Nair","3345"), ("USR-006","Kavya Reddy","6678"), ("USR-007","Arjun Kumar","1123"), ("USR-008","Sneha Patel","8856"), ("USR-009","Sunita Verma","4490"), ("USR-010","Rahul Joshi","7723"), ("USR-011","Deepika Rao","2234"), ("USR-012","Manish Iyer","5567"), ("USR-013","Pooja Pillai","8890"), ("USR-014","Nikhil Bose","1112"), ("USR-015","Ananya Das","4445"), ("USR-016","Karthik Menon","7778"), ("USR-017","Revathi Nair","2221"), ("USR-018","Siddharth Shah","5554"), ("USR-019","Lavanya Iyer","8887"), ("USR-020","Aditya Verma","1120"), ("USR-021","Meera Krishnan","4453"), ("USR-022","Tarun Agarwal","7786"), ("USR-023","Ishaan Chopra","2219"), ("USR-024","Riya Malhotra","5552"), ("USR-025","Varun Tiwari","8885"), ("USR-026","Nisha Pillai","1118"), ("USR-027","Gaurav Pandey","4447"), ("USR-028","Shweta Mishra","7780"), ("USR-029","Rahul Srivastava","2213"),("USR-030","Pallavi Desai","5546"), ("USR-031","Kunal Jain","8879"), ("USR-032","Preeti Yadav","1104"), ("USR-033","Vivek Khanna","4433"), ("USR-034","Smita Banerjee","7766"), ("USR-035","Harish Dubey","2205"), ("USR-036","Divya Kapoor","5538"), ("USR-037","Suresh Nambiar","8871"), ("USR-038","Tanya Bajaj","1096"), ("USR-039","Mohit Soni","4419"), ("USR-040","Ritika Ahuja","7752"), ("USR-041","Ajay Chauhan","2197"), ("USR-042","Bhavna Thakur","5530"), ("USR-043","Vinod Saxena","8863"), ("USR-044","Jyoti Rastogi","1088"), ("USR-045","Naveen Rao","4405"), ("USR-046","Poonam Shukla","7738"), ("USR-047","Sandeep Ghosh","2189"), ("USR-048","Alka Mehra","5522"), ("USR-049","Rajesh Kumar","8855"), ("USR-050","Sundar Rajan","1080"), ] # --------------------------------------------------------------------------- # City reference data (Indian + international for fraud rows) # --------------------------------------------------------------------------- CITIES = { "Mumbai": ("India", 19.076, 72.877), "Delhi": ("India", 28.613, 77.209), "Bengaluru": ("India", 12.971, 77.594), "Hyderabad": ("India", 17.385, 78.486), "Chennai": ("India", 13.082, 80.270), "Kolkata": ("India", 22.572, 88.363), "Pune": ("India", 18.520, 73.856), "Ahmedabad": ("India", 23.022, 72.571), # International (used only in fraud rows) "London": ("UK", 51.507, -0.127), "Singapore": ("Singapore", 1.352, 103.819), "Dubai": ("UAE", 25.204, 55.270), } DOMESTIC_CITIES = ["Mumbai","Delhi","Bengaluru","Hyderabad","Chennai","Kolkata","Pune","Ahmedabad"] MERCHANTS = [ "Swiggy","Zomato","Amazon","Flipkart","BigBasket","DMart", "BookMyShow","MakeMyTrip","IRCTC","Ola","Uber","Rapido", "Cafe Coffee Day","McDonald's","Domino's","KFC", "Reliance Digital","Croma","Apple Store", "Gold Palace Jewellers","SBI ATM","HDFC ATM", "Apollo Pharmacy","Myntra","Nykaa","Ajio", ] def random_amount(merchant, base_spend): """Amounts vary by merchant category.""" if "ATM" in merchant: return round(random.choice([2000,5000,10000,20000]), 2) if "Jewellers" in merchant: return round(random.uniform(15000, 150000), 2) if merchant in ["Amazon","Flipkart","Reliance Digital","Croma"]: return round(random.uniform(500, 25000), 2) if merchant in ["Swiggy","Zomato","Domino's","KFC","McDonald's","Cafe Coffee Day"]: return round(random.uniform(150, 1200), 2) return round(random.uniform(base_spend * 0.3, base_spend * 2.5), 2) def make_timestamp(base_dt, offset_minutes): return (base_dt + timedelta(minutes=offset_minutes)).strftime("%Y-%m-%d %H:%M:%S") # --------------------------------------------------------------------------- # Build the transaction list # --------------------------------------------------------------------------- transactions = [] txn_id_counter = 1 fraud_count = 0 # Assign each user a home city and typical spend range user_home = {u[0]: random.choice(DOMESTIC_CITIES) for u in USERS} user_avg = {u[0]: random.uniform(600, 4000) for u in USERS} base_date = datetime(2025, 8, 1, 8, 0, 0) # 1900 legitimate transactions for _ in range(1900): user_id, name, card = random.choice(USERS) city = user_home[user_id] country, lat, lon = CITIES[city] merchant = random.choice(MERCHANTS) amount = random_amount(merchant, user_avg[user_id]) offset = random.randint(0, 43200) # spread over 30 days transactions.append({ "transaction_id": f"TXN-{txn_id_counter:05d}", "user_id": user_id, "user_name": name, "card_last4": card, "amount": f"{amount:.2f}", "merchant": merchant, "city": city, "country": country, "lat": f"{lat:.4f}", "lon": f"{lon:.4f}", "timestamp": make_timestamp(base_date, offset), "is_fraud": "false", }) txn_id_counter += 1 # 2a. Impossible Travel fraud (~40 rows) intl_cities = ["London", "Singapore", "Dubai"] fraud_users = ["USR-007","USR-015","USR-023","USR-031","USR-039","USR-042", "USR-048","USR-003","USR-011","USR-019"] for fu in fraud_users: user_id, name, card = next(u for u in USERS if u[0] == fu) home_city = user_home[user_id] h_country, h_lat, h_lon = CITIES[home_city] intl = random.choice(intl_cities) i_country, i_lat, i_lon = CITIES[intl] offset = random.randint(5000, 40000) # Legitimate transaction at home transactions.append({ "transaction_id": f"TXN-{txn_id_counter:05d}", "user_id": user_id, "user_name": name, "card_last4": card, "amount": f"{random_amount('Swiggy', user_avg[user_id]):.2f}", "merchant": "Swiggy", "city": home_city, "country": h_country, "lat": f"{h_lat:.4f}", "lon": f"{h_lon:.4f}", "timestamp": make_timestamp(base_date, offset), "is_fraud": "false", }) txn_id_counter += 1 # Impossible: same card in international city 30-45 min later fraud_offset = offset + random.randint(30, 45) transactions.append({ "transaction_id": f"TXN-{txn_id_counter:05d}", "user_id": user_id, "user_name": name, "card_last4": card, "amount": f"{random.uniform(8000, 40000):.2f}", "merchant": random.choice(["Harrods","Luxury Retail","Duty Free","Casino Royale"]), "city": intl, "country": i_country, "lat": f"{i_lat:.4f}", "lon": f"{i_lon:.4f}", "timestamp": make_timestamp(base_date, fraud_offset), "is_fraud": "true", }) txn_id_counter += 1 fraud_count += 1 # 2b. Unusual amount fraud (~30 rows) amount_fraud_users = ["USR-002","USR-010","USR-018","USR-026","USR-034", "USR-004","USR-012","USR-020","USR-028","USR-036", "USR-044","USR-050","USR-006","USR-014","USR-022"] for fu in amount_fraud_users: user_id, name, card = next(u for u in USERS if u[0] == fu) city = user_home[user_id] country, lat, lon = CITIES[city] spike = round(user_avg[user_id] * random.uniform(12, 40), 2) offset = random.randint(2000, 38000) transactions.append({ "transaction_id": f"TXN-{txn_id_counter:05d}", "user_id": user_id, "user_name": name, "card_last4": card, "amount": f"{spike:.2f}", "merchant": random.choice(["Gold Palace Jewellers","Luxury Watch Store","Apple Store","Rolex Boutique"]), "city": city, "country": country, "lat": f"{lat:.4f}", "lon": f"{lon:.4f}", "timestamp": make_timestamp(base_date, offset), "is_fraud": "true", }) txn_id_counter += 1 fraud_count += 1 # 2c. Rapid fire fraud — 4 txns within 90 seconds (~17 users × 4 txns) rapid_users = ["USR-005","USR-013","USR-021","USR-029","USR-037", "USR-045","USR-008","USR-016","USR-024","USR-032", "USR-040","USR-001","USR-009","USR-017","USR-025", "USR-033","USR-041"] for fu in rapid_users: user_id, name, card = next(u for u in USERS if u[0] == fu) city = user_home[user_id] country, lat, lon = CITIES[city] offset = random.randint(3000, 35000) rapid_merchants = random.sample(MERCHANTS, 4) for i, m in enumerate(rapid_merchants): sec_offset = offset + (i * 22) / 60.0 # 22 seconds apart transactions.append({ "transaction_id": f"TXN-{txn_id_counter:05d}", "user_id": user_id, "user_name": name, "card_last4": card, "amount": f"{random.uniform(200, 800):.2f}", "merchant": m, "city": city, "country": country, "lat": f"{lat:.4f}", "lon": f"{lon:.4f}", "timestamp": make_timestamp(base_date, sec_offset), "is_fraud": "true" if i > 0 else "false", }) txn_id_counter += 1 fraud_count += 3 # Sort chronologically transactions.sort(key=lambda x: x["timestamp"]) # Renumber IDs after sort for i, t in enumerate(transactions, 1): t["transaction_id"] = f"TXN-{i:05d}" # Write CSV fieldnames = ["transaction_id","user_id","user_name","card_last4", "amount","merchant","city","country","lat","lon", "timestamp","is_fraud"] with open("transactions.csv", "w", newline="") as f: writer = csv.DictWriter(f, fieldnames=fieldnames) writer.writeheader() writer.writerows(transactions) legitimate = len(transactions) - fraud_count print(f"✅ Generated {len(transactions)} transactions → transactions.csv") print(f" Legitimate : {legitimate}") print(f" Fraud planted : {fraud_count}") print(f" Fraud rate : {fraud_count/len(transactions)*100:.1f}%") print() print("Run next: python transaction_producer.py")