https://github.com/kubosis/torch-rating

Seamless integration of sport rating systems into graph neural networks in the PyTorch environment
https://github.com/kubosis/torch-rating

gnn-model pytorch sports-analytics

Last synced: 3 months ago
JSON representation

Seamless integration of sport rating systems into graph neural networks in the PyTorch environment

• Host: GitHub
• URL: https://github.com/kubosis/torch-rating
• Owner: kubosis
• License: mit
• Created: 2023-10-19T12:09:32.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2024-12-06T10:23:59.000Z (10 months ago)
• Last Synced: 2025-06-16T19:17:15.824Z (4 months ago)
• Topics: gnn-model, pytorch, sports-analytics
• Language: Python
• Homepage:
• Size: 9.89 MB
• Stars: 4
• Watchers: 1
• Forks: 1
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
![Flake8 Status](https://github.com/kubosis/torch-rating/actions/workflows/quality.yml/badge.svg)

[![PyPI version](https://img.shields.io/pypi/v/torch-rating.svg)](https://pypi.org/project/torch-rating/)

![NeRa LOGO](https://github.com/kubosis/torch-rating/blob/main/docs/logo3.png?raw=true)

PyTorch based package for incorporating rating systems to neural networks. This package provides model rating layers. The resulting RatingRGNN can be found [here](https://github.com/kubosis/rating_gnn)

### Prerequisities

```

Python >= 3.10

```

### Installation

```commandline

pip install --upgrade pip

pip install torch-rating

```

### Nera - Neural rating

This package implements seamless integration of statistical rating systems into graph neural network in the PyTorch environment.

This project was developed as my Bachelor's thesis.

### Implemented rating layers and recurrent graph neural network architectures

- Elo rating

- Berrar rating

- Pi rating

![RatingRGNN architecture](https://raw.githubusercontent.com/kubosis/torch-rating/2d80c8f9b6f3faaf0a5a8e1d9cbecc8c7a2f44f3/docs/img/ratingRGNN.svg)

### Showcases of predictive validation accuracy on collected datasets:

Note: the RatingRGNN was fine-tuned only on the NBL dataset and then applied across the other.

![RatingRGNN architecture](https://github.com/kubosis/torch-rating/blob/main/docs/img/validation.png?raw=true)

Note: the accuracy is across time snapshots. These snapshots represent seasons. They do not represents epochs of iterating the whole dataset. The training was done only for one epoch.

![RatingRGNN architecture](https://github.com/kubosis/torch-rating/blob/main/docs/img/train_val_acc.png?raw=true)