Skip to content

TRANSACTION_DATA.md

Latest commit

 

History

History
196 lines (145 loc) · 7.3 KB

TRANSACTION_DATA.md

File metadata and controls

196 lines (145 loc) · 7.3 KB

Medici Bank Historical Transaction Data

Overview

This directory contains approximately 80,000 historically-themed banking transactions generated based on actual events from the Medici Bank's operations during 1390-1440. The data represents major financial and political events in Renaissance Italy.

The dataset was expanded from the original 20,000 transactions by running generate_additional_data.py, which:

  • Appended 60,000 more legitimate transactions (same historical distribution)
  • Injected a hidden embezzlement trail (~230 transactions, ~95,610 florins, 1420–1424) for forensic analysis exercises

Files

Generated Data Files

  • medici_transactions.csv (~11 MB) - Transaction data in CSV format (~80,000 rows)
  • medici_transactions.json (~29 MB) - Transaction data in JSON format (~80,000 records)

Scripts

  • generate_historical_data.py - Script to generate the initial 20,000 historical transactions
  • generate_additional_data.py - Script to expand to 80,000+ transactions and inject the embezzlement trail
  • validate_transactions.py - Script to validate the generated data

Data Structure

Each transaction contains the following fields:

Field Type Description
id Integer Unique transaction identifier
date ISO Date Transaction date (YYYY-MM-DD)
branch String Branch location (Florence, Rome, Venice, etc.)
type String Transaction type (see types below)
counterparty String The other party in the transaction
description String Detailed transaction description
debit_account String Account to debit
debit_amount Decimal Amount to debit (in florins)
credit_account String Primary account to credit
credit_amount Decimal Amount to credit (in florins)
credit_account_2 String Secondary credit account (optional)
credit_amount_2 Decimal Secondary credit amount (optional)
currency String Currency used (typically "florin")

Transaction Types

The data includes the following transaction types based on historical banking operations:

  1. deposit (31.7%) - Customer deposits, especially from papal sources
  2. operating_expense (13.6%) - Daily operating costs (wages, rent, supplies, etc.)
  3. war_financing (13.1%) - Loans to Florence, Venice for various wars
  4. loan_repayment (10.3%) - Loan repayments with interest
  5. loan_issuance (9.2%) - Loans to merchants and nobles
  6. bill_of_exchange (8.0%) - International money transfers (Medici innovation)
  7. withdrawal (7.8%) - Customer withdrawals
  8. alum_trade (6.3%) - Trade in alum from papal mines
  9. ransom_payment (<0.1%) - Special: Council of Constance ransom (35,000 florins)

Historical Context

The data is based on real historical events:

Western Schism and Papal Banking (1402-1420s)

  • In 1410, Pope John XXIII appointed the Medici as papal bankers
  • The Rome branch held ~100,000 florins in papal deposits
  • This was the bank's most important client relationship

Council of Constance (1414-1418)

  • The data includes the famous 35,000 florin ransom payment (May 29, 1415)
  • Giovanni di Bicci de' Medici paid this to secure Pope John XXIII's release
  • This was almost half the bank's profits from its first 20 years

Florentine-Milanese Wars

  • First War (1390-1402): Against Gian Galeazzo Visconti
  • Second War (1422-1426): Against Duke Filippo Maria of Milan
  • Significant war financing transactions during these periods

Wars in Lombardy (1423-1454)

  • Conflicts between Venice and Milan
  • Medici financed Florence's participation
  • Data shows increased war financing during this period

Branches Represented

  • Rome (33.1%) - Papal banking center
  • Florence (22.1%) - Home base
  • Venice (9.7%) - Major trading partner
  • London, Bruges, Avignon, Geneva, Milan - International network

Data Validation

All transactions maintain proper double-entry accounting:

  • Total Transactions: ~80,230
  • Every transaction is perfectly balanced (debits = credits).

Usage

Viewing the Data

# View first 10 transactions in CSV
head -11 medici_transactions.csv

# Count total transactions
wc -l medici_transactions.csv

# Search for specific events
grep "ransom" medici_transactions.csv
grep "war_financing" medici_transactions.csv

Generating New Data

# Generate fresh transaction data
python3 generate_historical_data.py

This will create new medici_transactions.csv and medici_transactions.json files.

Validating the Data

# Run validation checks
python3 validate_transactions.py

This validates:

  • CSV and JSON structure
  • Double-entry accounting (debits = credits)
  • Date formats
  • Transaction distributions
  • Historical event coverage

Sample Transactions

Council of Constance Ransom

date,branch,type,amount,description
1415-05-29,Constance,ransom_payment,35000.0,"Payment of 35,000 florin ransom for Pope John XXIII"

Papal Deposit

date,branch,type,amount,description
1390-01-05,Rome,deposit,38176.7,"Deposit from Vatican Treasury to Rome branch"

War Financing

date,branch,type,amount,description
1390-01-01,Florence,war_financing,82833.66,"Emergency war financing for Florence defense"

Loan with Interest

date,branch,type,debit,credit,credit2,description
1390-01-06,Florence,loan_repayment,829.16,720.31,108.85,"Loan repayment from Duke of Milan with interest"

Historical Accuracy

The data is designed to reflect:

  • ✓ Realistic transaction amounts (exponential distribution)
  • ✓ Historical interest rates (8-25% per annum)
  • ✓ Branch distribution (Rome dominant for papal banking)
  • ✓ Seasonal patterns in business activity
  • ✓ War periods showing increased financing
  • ✓ Major historical events (Council of Constance ransom)
  • ✓ Banking innovations (bills of exchange)
  • ✓ Papal monopolies (alum trade)

Hidden Forensic Exercise

The dataset includes a deliberately embedded embezzlement scenario for use in data engineering and forensic analysis exercises. A fictitious supplier named "Ser Benedetto Forniture" appears in the Florence branch operating expense records between 1420 and 1424, representing ~230 fraudulent transactions totalling approximately 95,610 florins.

The fraud is designed to be plausible on casual inspection but detectable through simple statistical analysis, including:

  • Benford's Law analysis of first digits
  • Vendor concentration analysis
  • Round-number clustering
  • Payment frequency analysis

For instructors: Full details of the scheme, detection methods, discussion questions, and grading rubric are in INSTRUCTOR_EMBEZZLEMENT_GUIDE.md. Do not distribute to students before the exercise.

Integration with Existing Code

This data can be imported into the existing medici-banking.py ledger system for demonstration purposes. Each transaction follows the double-entry accounting principles implemented in the main codebase.

Sources

The historical context is drawn from:

  • Raymond de Roover's "The Rise and Decline of the Medici Bank" (1963)
  • Contemporary chronicles and papal records
  • Historical research on Renaissance Italian banking

License

This data is part of the MediciMess educational project and is released under the MIT License.