DotaBrain

A Dota2 Hero Recommendation Engine Based On MachineLearning Techs And MonteCarlo Search

Introduction

DotaBrain is a dota2 game hero recommendation engine using machine learning and artificial intelligence technology.
DotaBrain learns a predictive model that maps the hero composition of both team to the match outcome, the predictive accuracy exceeds that of many experienced players in a test.
Based on this predictive model, DotaBrain further uses MonteCarlo Tree Search algorithm which is a key algorithm for many board games AI like AlphaGo to provide players with real-time hero recommendation.

Framework

Requirements

1.flask
2.scikit-learn
3.numpy

Getting Started

Install

git clone https://github.com/weiyu322/DotaBrain.git

Start Web App

cd DotaBrain/
python app.py

Using APIs

The web server provides two APIs: prediction API and recommendation API

prediction api: given the full hero composition of a match(10 heroes), return the predicitve result of the match

request form

POST /api/v1.0/predict HTTP/1.1
Content-type: application/json
Host: localhost:5000

{
  "radiant": ["Anti-Mage","Axe","Bane","Bloodseeker","Crystal Maiden"],

"dire": ["Drow Ranger","Earthshaker","Juggernaut","Mirana","Morphling"] }

* response form

HTTP/1.1 200 OK Date: Thu, 12 Jan 2017 08:34:15 GMT Content-Type: application/json Server: Werkzeug/0.11.4 Python/2.7.12

{ "radiantWinRate": 0.48737193751218433, "direWinRate": 0.51262806248781567 }

* recommendation api: given part of hero composition of a match(< 10 heroes), return topK hero recommendations
* request form

POST /api/v1.0/recommend HTTP/1.1 Content-type: application/json Host: localhost:5000

{ "ownSide": ["Anti-Mage","Axe","Bane"], "enemySide": ["Bloodseeker","Crystal Maiden","Drow Ranger"] "topK": 3 }

* response form