PDF Credit Card Statement Parser

Overview

This project is a PDF parser that extracts key data from credit card statements across multiple banks. It automatically detects high balances and potential fraud, calculates a risk score, and exports the results for review.

---

Features

• Extracts key data:
  • Card Member
  • Card Number (last 4 digits)
  • Card Variant
  • Statement Period
  • Payment Due Date
  • Total Amount Due / Balance
• Flags high balances (> ₹20,000)
• Detects potential fraud if Customer Identification Number (CIN) is missing
• Computes risk score and severity (Low / Medium / High)
• Exports results to CSV for easy review

---

How It Works

1. Reads PDF statements from a specified folder using pdfplumber.
2. Uses regex to extract relevant fields.
3. Flags high balances and missing Customer IDs.
4. Calculates a numeric risk score:
   • High Balance = 1 point
   • Missing Customer ID = 2 points
5. Maps total points to a risk severity:
   • 0 → Low
   • 1 → Medium
   • ≥2 → High
6. Outputs all parsed data and flags into a CSV file.

---

Tech Stack / Libraries

• Python 3.x – Core programming language
• pdfplumber – Extract text from PDFs
• re (regex) – Parse key fields from text
• pandas – Data manipulation and CSV export
• Optional visualization: matplotlib / seaborn / Plotly

---

Setup Instructions

1. Clone the repository:

   git clone https://github.com/anugavstudy/pdf-parser.git
   cd pdf-parser

2. Install dependencies:

   ```bash
   pip install pdfplumber pandas
   ```
3. Place your credit card PDF statements in the `statements/` folder.
4. Run the parser:

   ```bash
   python main.py
   ```
5. Check the output CSV:

   ```
   parsed_statements_with_risk.csv
   ```

---

## Example Output

| Card Member | Card Last 4 | Balance | High Balance Flag | Fraud Flag | Risk Points | Risk Severity |
| ----------- | ----------- | ------- | ----------------- | ---------- | ----------- | ------------- |
| ABC PQR     | 1234        | 24680   | True              | False      | 1           | Medium        |
| XYZ LMN     | 5678        | 8000    | False             | True       | 2           | High          |

---

## Why It Matters

* Automates tedious manual extraction of statements
* Detects potential fraud quickly and accurately
* Provides risk scoring to prioritize high-risk statements

---

## Future Enhancements

* Extract detailed transactions for fraud analysis
* Build dashboards for visualizing high-risk statements
* Implement automated email alerts for flagged statements

---