Transaction Frauds

This is inspired by JP Morgan Cybersecurity Study Case via Forage. The dataset named 'transaction.csv can be found in Kaggle.

Scenario

You are a cybersecurity analyst at the one of the largest financial companies in the world. Your job is to analyze a large dataset of fraud in Financial Payment Services. The dataset has five types of transactions:

• CASH-IN is any deposit.
• CASH-OUT is any withdrawal.
• DEBIT is a specific type of withdrawal in which the money is sent to the user’s bank account.
• PAYMENT is the purchase of goods or services.
• TRANSFER involves moving money from one user’s account to another user’s account.

Process

1. Read the dataset (transactions.csv) as a Pandas dataframe. Note that the first row of the CSV contains the column names.
2. Return the column names as a list from the dataframe.
3. Return the first k rows from the dataframe.
4. Return a random sample of k rows from the dataframe.
5. Return the Origin account balance delta v. Destination account balance delta scatter plot for Cash Out transactions (Source Delta & Delta Destination).
6. Return Fraud transactions that are flagged as frauds and how many of them are real frauds.

Implementation

• Transaction types bar chart:

• Transaction types frequencies:

• Delta Source:

• Fraud Detection:

Future Works:

• I would like to use free open source Facebook, Twitter, etc. scrappers, to gather the data and put them into csv extension file.
• More analysis can be conducted for this type of report.