https://github.com/offchan42/honnet

🤖 Deep learning predicts who would win in a HoN match given heroes of both team
https://github.com/offchan42/honnet

convolutional-neural-networks deep-learning game heroes-of-newerth hon keras neural-networks

Last synced: 2 months ago
JSON representation

🤖 Deep learning predicts who would win in a HoN match given heroes of both team

• Host: GitHub
• URL: https://github.com/offchan42/honnet
• Owner: offchan42
• License: apache-2.0
• Created: 2017-02-08T14:05:29.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2024-03-02T00:01:20.000Z (10 months ago)
• Last Synced: 2024-03-03T00:39:08.524Z (10 months ago)
• Topics: convolutional-neural-networks, deep-learning, game, heroes-of-newerth, hon, keras, neural-networks
• Language: Jupyter Notebook
• Homepage: https://github.com/off99555/HoNNeT-web
• Size: 5.54 MB
• Stars: 3
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# HoNNeT

Deep learning predicts who would win in a HoN match given heroes of both team.

[![HoNNeT Website Screenshot](Screenshot_12.png)](https://github.com/offchan42/HoNNeT-web)

The idea is that we will download match information in the past and train a deep learning model

which receives hero picks of both team as input then predict which team is winning.

**Note**: Heroes of Newerth (HoN) is a multiplayer online battle arena (MOBA) game similar to DOTA

but is very popular among Thai people.

## Role of each file

- `hon_db_retriever.ipynb` contains data preparation/wrangling process, this is

  probably the most time consuming part of the pipeline, it downloads HoN

  matches and save it into a **pickle** file. The data retrieval part is based on

  international HoN's API which is very limited in number of matches you can download per API call.

- `hon_matches.pkl` contains downloaded 100,000 matches

- `hon_db_analyzer.ipynb` contains deep learning part and data

  exploration/visualization

- `honnet_loader.ipynb` loads the trained model and save it for later use in

  multiple formats

- Other files are explained at

  the [web repository](https://github.com/offchan42/HoNNeT-web) because they are

  the same.

## Disclaimer

The data used for training the model needs to be updated regularly because heroes mechanisms are updated often.

Look at the modification date of the `honnet_brain.h5` to see when it was last updated.

**PS.** This repository is for gathering matches data and training deep learning model only.

For the user interface which is implemented as a web application on Heroku, you need to see

[HoNNeT-web](https://github.com/offchan42/HoNNeT-web) repository.