RetailBanking/description.md

Retail Banking Data Challenge Overview

Welcome to the retail banking data challenge! This dataset contains information for two predictive modeling tasks involving customer financial behavior, transaction monitoring, and credit risk assessment.

Directory & File Structure

The dataset is organized into CSV files and JSON logs:

├── customers_all.csv
├── accounts_all.csv
├── devices_all.csv
├── device_sessions_all.json
├── transactions_train.csv
├── transactions_test.csv
├── customer_panel_train.csv
├── customer_panel_test.csv

Customer Demographics

File: customers_all.csv

Columns:

• CustomerID: Unique identifier for each customer (e.g., C000001, C000002...)
• Age: Age of the customer in years
• Tenure: Number of years as a bank customer
• CreditScore: Credit score of the customer (300-850 scale)
• HomeCity: Customer's primary city of residence
• AnnualSalary: Customer's annual salary in USD

Account Information

File: accounts_all.csv

Columns:

• CustomerID: Identifier linking the account to a customer
• AccountID: Unique identifier for each account
• Type: Account type (checking, savings, credit_card)
• Balance: Current account balance in USD
• Limit: Credit limit for credit card accounts (empty for checking/savings)
• OpenDate: Date when the account was opened (YYYY-MM-DD format)

Device Associations

File: devices_all.csv

Columns:

• CustomerID: Identifier linking the device to a customer
• DeviceID: Unique identifier for each device used by the customer

Device Session Logs

File: device_sessions_all.json

Structure:

[
  {
    "CustomerID": "C000001",
    "SessionID": "551a1ac3-4f20-4ce9-b6d6-c7cb07b460d5",
    "Timestamp": "2025-01-01T01:11:12",
    "City": "City096",
    "IP": "192.168.1.100",
    "DeviceID": "B5E74E",
    "Actions": [
      "login",
      "account_view",
      "payment",
      "logout"
    ]
  }
]

Contains detailed session information including device usage patterns, location data, and user actions performed during each banking session.

Transaction Data

Files: transactions_train.csv, transactions_test.csv

Columns:

• Timestamp_dt: Date portion of the transaction timestamp
• TxnID: Unique identifier for each transaction
• Timestamp: Full timestamp of the transaction (YYYY-MM-DD HH:MM:SS)
• SessionID: Identifier linking to device session (if online transaction)
• CustomerID: Identifier linking the transaction to a customer
• SrcAccount: Source account identifier
• DstAccount: Destination account identifier (for transfers)
• Channel: Transaction channel (online, ATM, branch)
• MCC_Group: Merchant category code group (Groceries, Restaurants, Online Retail, etc.)
• Amount: Transaction amount in USD
• FraudLabel (train only): Binary indicator (0 or 1) of whether the transaction is fraudulent (target variable)

Note: FraudLabel is not provided in the test set.

Customer Panel Data

Files: customer_panel_train.csv, customer_panel_test.csv

Columns:

• CustomerID: Identifier linking to customer information
• Week: Week number (1-26 representing weeks in the 6-month period)
• Utilisation: Credit utilization ratio (credit card balance / credit limit)
• PaymentRatio: Ratio of payments made to previous week's credit card balance
• HardInquiries: Number of hard credit inquiries during the period
• DefaultLabel (train only): Binary indicator (0 or 1) of whether the customer will default within the period (target variable)

Note: DefaultLabel is not provided in the test set.

Challenge Tasks & Submission Format

Challenge 1: Fraud Detection

Predict the FraudLabel for each transaction in transactions_test.csv.

Submission CSV: TxnID,FraudLabel

Evaluation Metric: Macro-F1.

Challenge 2: Credit Default Prediction

Predict the DefaultLabel for each customer-week combination in customer_panel_test.csv.

Submission CSV: CustomerID,Week,DefaultLabel

Evaluation Metric: Macro-F1.

Notes & Tips

• Only the described columns are provided. Participants must infer any latent variables from provided transaction data, device sessions, and behavioral patterns.
• Data integration across multiple files using CustomerID is essential for effective modeling.
• Ensure submissions strictly adhere to the specified CSV formats.

Good luck!