Skip to content
Credit Card Fraud
Credit Card Fraud
This dataset consists of credit card transactions in the western United States. It includes information about each transaction including customer details, the merchant and category of purchase, and whether or not the transaction was a fraud.
import pandas as pd
df = pd.read_csv('credit_card_fraud.csv') Data Dictionary
| transdatetrans_time | Transaction DateTime |
|---|---|
| merchant | Merchant Name |
| category | Category of Merchant |
| amt | Amount of Transaction |
| city | City of Credit Card Holder |
| state | State of Credit Card Holder |
| lat | Latitude Location of Purchase |
| long | Longitude Location of Purchase |
| city_pop | Credit Card Holder's City Population |
| job | Job of Credit Card Holder |
| dob | Date of Birth of Credit Card Holder |
| trans_num | Transaction Number |
| merch_lat | Latitude Location of Merchant |
| merch_long | Longitude Location of Merchant |
| is_fraud | Whether Transaction is Fraud (1) or Not (0) |
Source of dataset. The data was partially cleaned and adapted by DataCamp.
Exploratory Analysis
I would like to investigate a few different things to start, such as:
- Locations where fraud is most likely to occur
- Distance between purchase location and location of merchant
- Date and Time
- Age of customer
# import necessary packages
import numpy as np
import math
import matplotlib.pyplot as plt
import seaborn as sns
# look for missing values
print(df.describe())
print(df.info())
#convert datetime values
pd.to_datetime(df['dob'])
pd.to_datetime(df['trans_date_trans_time'])# Fraud and number of transactions by state
fraud_by_state = df.groupby('state')['is_fraud'].mean().reset_index()
transactions_by_state = df.groupby('state').size()
# Fraud and number of transactions by city
fraud_by_city = df.groupby('city')['is_fraud'].mean().reset_index().sort_values(by = 'is_fraud', ascending= False).head(20)
transactions_by_city = df.groupby('city').size()
#set sns style and color palette
sns.set_style('whitegrid')
sns.barplot(data = fraud_by_state, x = 'state',y='is_fraud',palette = 'Blues')
plt.title('Fraud Rate by State')
plt.show()
plt.figure(figsize=(6,12))
sns.barplot(data = fraud_by_city, y = 'city',x='is_fraud',orient='h', palette = 'Blues')
plt.title('Cities with Highest Fraud Rate')
plt.show()
#Isolate cities with 100% fraud rate to investigate
full_fraud_cities = fraud_by_city[fraud_by_city['is_fraud']==1]
full_fraud_cities = full_fraud_cities['city']
print('Cities with 100% Fraud Rate')
display(full_fraud_cities)
display(df[df['city'].isin(full_fraud_cities)])# Fraud and number of transactions by category
fraud_by_category = df.groupby('category')['is_fraud'].mean().reset_index().sort_values(by='is_fraud',ascending=False,)
transactions_by_category = df.groupby('category').size()
#plot figure
plt.figure(figsize=(6,12))
sns.barplot(data = fraud_by_category, y = 'category',x='is_fraud',orient='h',palette = 'Blues')
plt.title('Fraud Level by Purchase Category')
plt.show()
# add those features to the dataset
#df = pd.merge(df, pd.concat([fraud_by_city,transactions_by_city], axis=1), on='city', how='left', suffixes=['rate','transaction_count'])# create a function to calculate distance between two points using Haversine formula
def haversine(lat1, lon1, lat2, lon2):
R = 6371 # radius of the earth in km
dLat = math.radians(lat2 - lat1)
dLon = math.radians(lon2 - lon1)
a = math.sin(dLat / 2) * math.sin(dLat / 2) + math.cos(math.radians(lat1)) * math.cos(math.radians(lat2)) * math.sin(dLon / 2) * math.sin(dLon / 2)
c = 2 * math.atan2(math.sqrt(a), math.sqrt(1 - a))
d = R * c
return d
# create a new feature 'distance' that calculates the distance between the purchase and merchant locations
df['distance'] = df.apply(lambda row: haversine(row['lat'], row['long'], row['merch_lat'], row['merch_long']), axis=1)
display(df.head(25))# Correlation Matrix
corr = df.corr()
sns.heatmap(corr,cmap='Blues')
plt.show()
Identifying Customers
There is no sspecific customer identification in the dataset, but between the birthdate, city, and job title, it should be possible to parse out individual customers to be able to identify unique behavior to that customer that might make it easier for models to spot outliers.
#Identify unique customers based on a combination of birthdate, city_population, and job
df['customer_id_test'] = df['dob'].astype(str)+df['city'].astype(str)+df['job'].astype(str)
customers = df.groupby('customer_id_test')
# Get the unique customers and create and apply customer_ids
unique_customers = list(customers.groups.keys())
df["customer_id"] = df["customer_id_test"].map(dict(zip(unique_customers, range(len(unique_customers)))))
display(df.head(25))
#test on one customer_id to make sure the id lines up
display(df[df['customer_id']==122])
Unique Customer Behavior
Now that I have each customer uniquely identified, and assigned a customer id, I would like to map out customer behavior specific to each customer such as
- 'avg_purchase_price'
- 'avg_distance'
I would also like to include the quantiles for each