Transaction parsing and merchant enrichment pipeline for Google Pay data.
PayParse is an early-stage prototype focused on processing raw payment history and transforming it into structured, analyzable data.
The project emphasizes:
- Data parsing and cleaning
- Feature extraction from transaction records
- Merchant enrichment using external APIs
- Preparing a foundation for future machine learning workflows
- Parse Google Pay transaction history exports
- Clean and normalize raw transaction data
- Extract time-based features (hour, day, month)
- Enrich merchant data using Google Places API
- Perform heuristic-based categorization
- Python
- FastAPI (backend structure)
- Google Places API
- RapidFuzz (string matching)
This project currently focuses on data processing and enrichment.
It does not yet include a trained ML model or predictive system.
Planned improvements:
- Introduce a defined prediction task (e.g., spending classification or forecasting)
- Build training and evaluation pipeline
- Add model-based recommendations
- Improve generalization beyond single-user data
- Data ingestion → parsing and cleaning
- Feature engineering → timestamp-based features
- Merchant enrichment → API + fuzzy matching
- Output → structured dataset for analysis
- No machine learning model yet
- Limited dataset scope
- Depends on external API for enrichment
- No authentication or production deployment
- Explore financial data processing pipelines
- Serve as a base for future ML experimentation
- Understand real-world data cleaning and enrichment challenges
This is an experimental project and is actively being improved.