causalml-create-credit-card-data.py

import numpy as np
import pandas as pd
import csv
import random
import string

### Create full dataset of 5 million credit card customers ###

def generate_cust_id():
    return ''.join(random.choices(string.ascii_uppercase + string.digits, k=7))

def generate_age():
    return random.randint(25, 35)

def generate_num_credit_lines():
    return random.randint(1, 3)

def generate_income_tier():
    return random.randint(1, 3)

def generate_spend_grocery():
    return round(random.uniform(0.1, 0.6), 2)

def generate_spend_gas():
    return round(random.uniform(0.1, 0.6), 2)

def generate_spend_dineout():
    return round(random.uniform(0.1, 0.6), 2)

def generate_spend_travel():
    return round(random.uniform(0, 0.2), 2)

def generate_balance():
    return round(random.uniform(500, 5000), 2)

def generate_balance_frequency():
    return round(random.uniform(0.2, 0.8), 2)

def generate_avg_trn_value():
    return random.randint(10, 30)

def generate_purchases():
    return round(random.uniform(1, 100), 2)

def generate_oneoff_purchases():
    return round(random.uniform(0, 1), 2)

def generate_installments_purchases(oneoff_purchases):
    return round(1 - oneoff_purchases, 2)

def generate_cash_advances():
    return random.randint(0, 5)

def generate_forex_trn():
    return random.randint(0, 10)

def calculate_credit_limit(balance, purchases, avg_trn_value):
    return balance + purchases * avg_trn_value

def generate_payments():
    return random.randint(6, 12)

def generate_prcfullpayment():
    return round(random.uniform(0.5, 1), 2)

with open('synthetic_credit_card_dataset.csv', mode='w', newline='') as file:
    writer = csv.writer(file)
    
    writer.writerow(['CUST_ID', 'AGE', 'NUM_CREDIT_LINES', 'INCOME_TIER', 'SPEND_GROCERY', 'SPEND_GAS', 'SPEND_DINEOUT', 'SPEND_TRAVEL', 'BALANCE', 'BALANCE_FREQUENCY', 'AVG_TRN_VALUE', 'PURCHASES', 'ONEOFF_PURCHASES', 'INSTALLMENTS_PURCHASES', 'CASH_ADVANCES', 'FOREX_TRN', 'CREDIT_LIMIT', 'PAYMENTS', 'PRCFULLPAYMENT', 'TENURE'])
    
    for i in range(5000000):
        oneoff_purchases = generate_oneoff_purchases()
        balance = generate_balance()
        balance_frequency = generate_balance_frequency()
        avg_trn_value = generate_avg_trn_value()
        
        spend_grocery = generate_spend_grocery()
        spend_gas = generate_spend_gas()
        spend_dineout = generate_spend_dineout()
        spend_travel = generate_spend_travel()
        
        spend_grocery /= (spend_grocery+spend_gas+spend_dineout+spend_travel)
        spend_gas /= (spend_grocery+spend_gas+spend_dineout+spend_travel)
        spend_dineout /= (spend_grocery+spend_gas+spend_dineout+spend_travel)
        spend_travel /= (spend_grocery+spend_gas+spend_dineout+spend_travel)
        
        writer.writerow([generate_cust_id(), 
                         generate_age(), 
                         generate_num_credit_lines(), 
                         generate_income_tier(), 
                         spend_grocery, 
                         spend_gas, 
                         spend_dineout, 
                         spend_travel, 
                         balance, 
                         balance_frequency, 
                         avg_trn_value, 
                         generate_purchases(), 
                         oneoff_purchases,
                         generate_installments_purchases(oneoff_purchases),
                         generate_cash_advances(),
                         generate_forex_trn(),
                         calculate_credit_limit(balance, 
                                                balance_frequency,
                                                avg_trn_value),
                         generate_payments(),
                         generate_prcfullpayment(),
                         10])


### Create pilot dataset using 1% of the customers ###

data = pd.read_csv('synthetic_credit_card_dataset.csv')

pilot = data.sample(frac=0.01, random_state=1)
pilot['Treated'] = np.random.choice([0,1], pilot.shape[0])

pilot['Treatment'] = pd.cut(pilot['PURCHASES'] * pilot['AVG_TRN_VALUE'], 3, labels=["Annual Fee x 30%","Annual Fee x 50%","Annual Fee x 70%"])
pilot['Treatment'] = np.where(pilot['Treated'], pilot['Treatment'], "Full Annual Fee")

pilot['Treated'] = np.random.choice([0,1], pilot.shape[0])
pilot['Renewed'] = np.where(
    pilot['Treated'], 
    np.where(pilot['Treatment']=="Annual Fee x 30%", np.random.binomial(1, 0.7, pilot.shape[0]), 
            np.where(pilot['Treatment']=="Annual Fee x 50%", np.random.binomial(1, 0.7, pilot.shape[0]), 
                    np.random.binomial(1, 0.7, pilot.shape[0]))), 
    np.random.binomial(1, 0.4, pilot.shape[0])
)

pilot.to_csv('pilot_credit_card.csv', index=False)