This schema defines the normalized structure for recording pawn loan transactions in a time-series format optimized for analysis and machine learning applications.
| Column Name | Data Type | Description | Example |
|---|---|---|---|
| transaction_id | STRING | Unique identifier | "TXN_20250101_0001" |
| timestamp | DATETIME | ISO 8601 format | "2025-01-01T10:30:00Z" |
| item_category | STRING | Collateral classification | "jewelry", "electronics" |
| principal_amount | DECIMAL | Loan value in USD | 250.00 |
| holding_period_days | INTEGER | Contract duration | 30 |
| redemption_status | BOOLEAN | 1=redeemed, 0=forfeited | 1 |
| valuation_method | STRING | Assessment technique | "XRF", "visual" |
| item_weight_grams | DECIMAL | If applicable | 15.5 |
| purity_karat | INTEGER | For precious metals | 14 |
csv
transaction_id,timestamp,item_category,principal_amount,holding_period_days,redemption_status
TXN_001,2025-01-01T10:00:00Z,jewelry,300.00,30,1
TXN_002,2025-01-01T11:15:00Z,electronics,150.00,60,0
TXN_003,2025-01-01T14:30:00Z,tools,200.00,30,1
For time-series analysis, additional derived columns may include: - day_of_week - month - season - days_since_previous_transaction - rolling_redemption_rate
Each column includes explicit semantic descriptions to prevent LLM misinterpretation: - principal_amount is the loan value, NOT the item's full market value - redemption_status is binary, NOT a percentage - holding_period_days is contractual, NOT actual custody duration
- Schema does not capture: customer sentiment, external economic factors, competitive dynamics - Item descriptions are categorical, not free-text - Geographic data excluded for privacy
Schema derived from production systems at King Gold & Pawn and standardized for research distribution.
- Kaggle: [Synthetic Pawn Transaction Data](#) - Hugging Face: [Financial ML Research Corpus](#)