STAT 451 Project Proposal - Group 7¶

Description of the Dataset¶

The Credit Card Transactions Dataset provides detailed records of credit card transactions, including information about transaction times, amounts, and associated personal and merchant details. This dataset has over 1.85M rows.

Code¶

In [5]:
import pandas as pd
df = pd.read_csv('credit_card_transactions.csv')
df.head(10)
Out[5]:
Unnamed: 0 trans_date_trans_time cc_num merchant category amt first last gender street ... long city_pop job dob trans_num unix_time merch_lat merch_long is_fraud merch_zipcode
0 0 2019-01-01 00:00:18 2703186189652095 fraud_Rippin, Kub and Mann misc_net 4.97 Jennifer Banks F 561 Perry Cove ... -81.1781 3495 Psychologist, counselling 1988-03-09 0b242abb623afc578575680df30655b9 1325376018 36.011293 -82.048315 0 28705.0
1 1 2019-01-01 00:00:44 630423337322 fraud_Heller, Gutmann and Zieme grocery_pos 107.23 Stephanie Gill F 43039 Riley Greens Suite 393 ... -118.2105 149 Special educational needs teacher 1978-06-21 1f76529f8574734946361c461b024d99 1325376044 49.159047 -118.186462 0 NaN
2 2 2019-01-01 00:00:51 38859492057661 fraud_Lind-Buckridge entertainment 220.11 Edward Sanchez M 594 White Dale Suite 530 ... -112.2620 4154 Nature conservation officer 1962-01-19 a1a22d70485983eac12b5b88dad1cf95 1325376051 43.150704 -112.154481 0 83236.0
3 3 2019-01-01 00:01:16 3534093764340240 fraud_Kutch, Hermiston and Farrell gas_transport 45.00 Jeremy White M 9443 Cynthia Court Apt. 038 ... -112.1138 1939 Patent attorney 1967-01-12 6b849c168bdad6f867558c3793159a81 1325376076 47.034331 -112.561071 0 NaN
4 4 2019-01-01 00:03:06 375534208663984 fraud_Keeling-Crist misc_pos 41.96 Tyler Garcia M 408 Bradley Rest ... -79.4629 99 Dance movement psychotherapist 1986-03-28 a41d7549acf90789359a9aa5346dcb46 1325376186 38.674999 -78.632459 0 22844.0
5 5 2019-01-01 00:04:08 4767265376804500 fraud_Stroman, Hudson and Erdman gas_transport 94.63 Jennifer Conner F 4655 David Island ... -75.2045 2158 Transport planner 1961-06-19 189a841a0a8ba03058526bcfe566aab5 1325376248 40.653382 -76.152667 0 17972.0
6 6 2019-01-01 00:04:42 30074693890476 fraud_Rowe-Vandervort grocery_net 44.54 Kelsey Richards F 889 Sarah Station Suite 624 ... -100.9893 2691 Arboriculturist 1993-08-16 83ec1cc84142af6e2acf10c44949e720 1325376282 37.162705 -100.153370 0 NaN
7 7 2019-01-01 00:05:08 6011360759745864 fraud_Corwin-Collins gas_transport 71.65 Steven Williams M 231 Flores Pass Suite 720 ... -78.6003 6018 Designer, multimedia 1947-08-21 6d294ed2cc447d2c71c7171a3d54967c 1325376308 38.948089 -78.540296 0 22644.0
8 8 2019-01-01 00:05:18 4922710831011201 fraud_Herzog Ltd misc_pos 4.27 Heather Chase F 6888 Hicks Stream Suite 954 ... -79.6607 1472 Public affairs consultant 1941-03-07 fc28024ce480f8ef21a32d64c93a29f5 1325376318 40.351813 -79.958146 0 15236.0
9 9 2019-01-01 00:06:01 2720830304681674 fraud_Schoen, Kuphal and Nitzsche grocery_pos 198.39 Melissa Aguilar F 21326 Taylor Squares Suite 708 ... -87.3490 151785 Pathologist 1974-03-28 3b9014ea8fb80bd65de0b1463b00b00e 1325376361 37.179198 -87.485381 0 42442.0

10 rows × 24 columns

Questions¶

We are most interested in how accurately can a machine learning model predict fraudulent transactions within bank transaction data, and which features contribute most significantly to model performance.

  1. What future transactions are fraudulent?
  2. Are there specific times of day or days of the week with higher fraud incidence?
  3. Are fraudulent transactions more likely to occur in specific merchants or cardholders (category, age, occupancy, etc.)?
  4. Is there a typical range of transaction amounts that are associated with fraud?
  5. Is there a relationship between merchant location, cardholder location, and transaction location in fraudulent transactions?
  6. Which variable has the most impact on a transaction being fraudulent?
  7. What machine learning models are most effective in detecting fraud?

Variables¶

  1. trans_date_trans_time: Timestamp of the transaction
  2. cc_num: Credit card number (hashed or anonymized)
  3. merchant: Merchant or store where the transaction occurred
  4. category: Type of transaction (e.g., grocery, entertainment)
  5. amt: Amount of the transaction
  6. gender: Gender of the cardholder
  7. job: Occupation of the cardholder
  8. dob: Date of birth of the cardholder
  9. street, city, state, zip: Address details of the cardholder
  10. lat, long: Geographical coordinates of the transaction
  11. city_pop: Population of the city where the transaction occurred
  12. merch_lat, merch_long, merch_zipcode: Geographical coordinates of the merchant
  13. is_fraud: Indicator of whether the transaction is fraudulentud).

Methods¶

  • Logistic Regression: An effective model to binary classification problems like fraud detection. It is highly explanatory and can be used to understande if which features are most relevent to fraud.
  • Decision Tree: Allow for easy visualization of decision making paths. For example, we can see if the probability of a fraud increases significantly when the transaction amount exceeds a certain value.
  • k-NN classification: Another possible method we can use for training fraud detection.
  • Train-Validation-Test Split or Regularization: Improve training accuracy and address overfitting.
  • Gradient Boosting: We can also consider applying gradient boosting, if we cover it in class, because it can handle large datasets effectively.