Deep learning predicts who would win in a HoN match given heroes of both team.
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.
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 matcheshon_db_analyzer.ipynb
contains deep learning part and data exploration/visualizationhonnet_loader.ipynb
loads the trained model and save it for later use in multiple formats- Other files are explained at the web repository because they are the same.
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 repository.