Data Mining Project Dota 2 Heroes Suggestion Nattapat Maleehuan 5410500199 Pakawat Kijsuwanpaisarn 5410500229 Teppakorn Thanutanad 5410501373
Data Mining ProjectDota 2 Heroes Suggestion
Nattapat Maleehuan 5410500199Pakawat Kijsuwanpaisarn 5410500229
Teppakorn Thanutanad 5410501373
http://www.gamersbook.com/Portals/0/images/2013/11/1_dota-2-screenshot-48.jpg
http://i.ytimg.com/vi/BnyDujra7yU/maxresdefault.jpghttp://i.imgur.com/JbCRo5F.jpg
StrategyGame
http://www.cyborgmatt.com/wp-content/uploads/2012/10/Dota2_CM_Picker.jpg
109 Hero
http://www.dotafire.com/dota-2/hero/
Difference
hero skill, status, power
http://www.google.com
http://www.google.com
Frustrated
Is there any best and also
trustable ways to draft the heroes?
Draft Hero Professionally
Input data
Match Detail
Parse HTML
10,893 records
4 HoursCrawler run
After dinner
Data Pre-processing
- Each hero will transform to number.
1 If Hero A is picked by my side
-1 If Hero A is picked by opponent’s side
0.3 If Hero A is banned by my side
-0.3 If Hero A is banned by opponent’s side
Transform hero to cost
Techniques➢ Neural Network
○ Pre-processing -> Data1, Data2, …, Data20
○ Datai = Data contain pick/ban hero at i times.○ Datai -> Neural Network -> Modeli
➢ Suggestion○ Modelj suggest hero at postion j
■ If Having pick/ban information j-1 heroes can predict win_rate from ANN(learning from Dataj).
At suggestion hero at phase 5,model need learn first 5 hero.
Greedy Searching is that finding the hero that get highest win rate.
Model Creation
Rapidminer
Parameter
Learning
Neural NetFrom previous parameter
Web Application
❏ http://61.91.69.73/datamining/pick.html
Test❏ Accuracy of Model 61%❏ 4 online-matchmaking games
❏ 4 wins 0 losses
WINGame 1
WINGame 2
WINGame 3
WINGame 4
Dota 2 Events
http://www.hardcoregamer.com/wp-content/uploads/2014/06/dota2-576x309.jpg
http://static.fragbite.se/stock/d99ddb0a003d25cb2a30.jpg
Thank You!