Fake Transaction Data Generator

Overview

Prototype Python utility for generating fictional financial transaction data for audit analytics practice, fraud/anomaly detection exercises, and data science education.

The generator creates two CSV files:

• fake_transactions.csv — generated transactions with normal and irregular records.
• irregularities.csv — a ledger of injected irregularities and descriptions.

Maturity note: this is an educational audit-analytics utility, not a production synthetic-data platform. It is intentionally small, standard-library only, and suited to examples, workshops, and detector testing.

Features

• Configurable transaction volume and date range.
• Recurring and random transaction generation.
• Benford-style first-digit amount distribution for normal random transactions.
• Configurable irregularities/anomalies.
• Separate irregularity ledger for validation and training labels.
• Optional --seed argument for reproducible generated datasets.
• Simple example analysis script for detecting selected irregularity patterns.

Requirements

• Python 3.10 or higher.
• Runtime: Python standard library only.
• Tests: pytest.

Installation

git clone https://github.com/busera/fake_transaction_data_generator.git
cd fake_transaction_data_generator
python -m pip install pytest

Usage

Generate the sample dataset:

python fake_generator.py -c config.json -o fake_transactions.csv -a irregularities.csv --seed 42

Arguments:

• -c / --config: path to the configuration file. Default: config.json.
• -o / --output: output CSV file for transactions. Default: fake_transactions.csv.
• -a / --anomalies: output CSV file for irregularities. Default: irregularities.csv.
• --seed: optional integer seed for reproducible output.

Run a basic detector example after generating the CSV:

python examples/analyze_irregularities.py fake_transactions.csv

Run the tests:

python -m pytest -v

Output Files

Transactions CSV

Columns:

• Transaction ID
• Date
• Type
• Amount
• Account
• Description
• Vendor

Irregularities CSV

Columns:

• Transaction ID
• Irregularity Type
• Description

The irregularities file acts as a label ledger for validating detection logic or training exercises.

Configuration

config.json controls the generated period, volume, vendors, recurring transactions, and irregularity counts.

Example structure:

{
  "num_transactions": 1000,
  "start_date": "2023-01-01",
  "end_date": "2023-12-31",
  "irregularities": {
    "high_amount": {"count": 10},
    "frequency_change": {"count": 10},
    "double_spend": {"count": 10},
    "missing_id": {"count": 2},
    "incorrect_date": {"count": 10},
    "mismatched_description": {"count": 10},
    "wrong_account": {"count": 10},
    "personal_expense": {"count": 10},
    "benford_violation": {"count": 10},
    "subtle_skimming": {"count": 30},
    "seasonal_anomaly": {"count": 10},
    "round_number_bias": {"count": 10},
    "cumulative_irregularity": {
      "enabled": true,
      "count": 36,
      "threshold": 0.005
    }
  },
  "vendors": ["ABC Office Supplies", "XYZ Tech Solutions", "123 Cleaning Services"],
  "personal_vendors": ["Luxury Resort & Spa", "Designer Clothing Co.", "Gourmet Restaurant"],
  "personal_expense_descriptions": ["Team Building Retreat", "Client Entertainment", "Office Decor"],
  "recurring_transactions": [
    {
      "vendor": "City Power & Utilities",
      "amount": 500,
      "day": 15,
      "description": "Monthly Utility Bill"
    }
  ]
}

Irregularity Types

• high_amount: unusually large transaction amount.
• frequency_change: altered date for a recurring transaction.
• double_spend: duplicate transaction with a new ID and timestamp offset.
• missing_id: transaction with no transaction ID.
• incorrect_date: future-dated transaction.
• mismatched_description: description conflicts with transaction type.
• wrong_account: invalid account-number pattern.
• personal_expense: personal vendor/description disguised as business expense.
• benford_violation: amount distribution biased away from Benford's Law.
• subtle_skimming: small reductions across nearby transactions.
• seasonal_anomaly: seasonal mismatch, such as summer equipment in winter.
• round_number_bias: suspiciously round transaction amounts.
• cumulative_irregularity: small cumulative increases across multiple transactions.

Purpose and Use Cases

This project is useful for:

• Audit analytics workshops and training.
• Testing anomaly-detection scripts.
• Demonstrating how labeled irregularities support validation.
• Practicing data analysis without exposing confidential financial records.

Limitations

• It does not model a full accounting subledger or ERP export.
• Amounts are floats because this is synthetic training data, not financial posting logic.
• The included detector example is intentionally simple and rule-based.
• Generated patterns are educational examples, not proof of real-world fraud typologies.

License

MIT License. See LICENSE.