README.md

---
license: gpl-3.0
task_categories:
  - tabular-classification
  - tabular-regression
tags:
  - payments
  - fintech
  - payment-routing
  - payment-intelligence
  - tabular
  - synthetic-data
  - machine-learning
  - catboost
  - open-source
pretty_name: PayMind Reference Dataset
---

# PayMind Reference Dataset

Synthetic/reference payment-routing data for **PayMind**, an open-source payment route intelligence engine.

This dataset is designed to demonstrate PayMind's training, evaluation, and routing workflow across route selection, transaction reliability, and expected settlement time.

> **Important:** This dataset contains synthetic/reference data only. It does not contain real customers, real transactions, payment credentials, personally identifiable information, or production payment-provider performance data.

## Dataset Files

The dataset contains three primary CSV files:

### `payment_method.csv`

Reference data for training the payment-method candidate model.

The model learns which payment routes are most relevant for a given transaction context.

### `success.csv`

Reference data for training the transaction-success model.

The model estimates the probability of a transaction succeeding for an eligible payment route.

### `arrival.csv`

Reference data for training the settlement model.

The model estimates expected transaction arrival/settlement times, including typical and more conservative settlement estimates.

## Tasks

PayMind uses the dataset across multiple tabular machine-learning tasks:

- **Payment route selection** — identify relevant payment methods for a transaction.
- **Transaction reliability** — estimate the probability of successful processing.
- **Settlement estimation** — estimate expected transaction arrival time.
- **Route intelligence** — combine model outputs with eligibility rules and estimated fees to rank available payment routes.

## Intended Use

This dataset is intended for:

- Demonstrating the PayMind architecture
- Training the PayMind reference models
- Testing the PayMind training pipeline
- Software development and integration testing
- Machine-learning experimentation
- Educational and research use

Users deploying PayMind in a real payment environment should train models using appropriately governed data from their own environment.

## Data Source

The data provided in this repository is **synthetic/reference data** created for the PayMind open-source project.

It should not be interpreted as observed behaviour of any real payment provider, financial institution, customer, or payment network.

## Privacy

The reference dataset is designed to contain no real:

- Customer identities
- Names or email addresses
- Account or card details
- Payment credentials
- Authentication tokens or API keys
- Production transaction records
- Personally identifiable information (PII)
- Proprietary payment-provider performance data

## Reference Models

The PayMind reference models are trained using synthetic/reference data and are intended to demonstrate the architecture rather than provide production payment-routing benchmarks.

**Models:**  
https://huggingface.co/navk8690/paymind-reference-models

## Live Demo

Try PayMind through the interactive Hugging Face Space:

https://huggingface.co/spaces/navk8690/paymind

The demo evaluates transaction context, eligible payment routes, predicted reliability, settlement expectations, and estimated cost to produce a ranked route recommendation.

## Source Code

PayMind is open source.

**GitHub:**  
https://github.com/navjotk8690/paymind

The repository includes the SDK, API, model implementations, training pipeline, data contracts, reference configuration, tests, and Gradio demo.

## Training

The PayMind project provides a training pipeline for rebuilding the reference models or training models using your own compatible datasets.

The expected datasets are:

```text
payment_method.csv
success.csv
arrival.csv
```

Refer to the PayMind repository documentation for the exact schemas and training instructions.

## Limitations

This dataset is provided as a reference implementation dataset.

Results obtained from models trained on this data:

- Do not represent actual payment-provider performance
- Should not be treated as production routing benchmarks
- Do not guarantee transaction success or settlement time
- Should not be used as a substitute for production-specific model validation
- May not reflect the distributions, constraints, or behaviour of a real payment environment

Production users should validate their own data, models, eligibility rules, fee configuration, and ranking strategy before deployment.

## License

This dataset is released under the **GNU General Public License v3.0 (GPL-3.0)** as part of the PayMind open-source project.

See the repository license for full terms.