Using Probabilistic Models and Data Mining Techniques in Online Computer Games

Please enable javascript in your browser.

Khosravinia, Sina | 2019

432 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 52947 (01)
University: Sharif University of Technology
Department: Industrial Engineering
Advisor(s): Haji, Babak
Abstract:
Computer video games have had a huge growth in recent years, and Dota 2 is one of the most popular video games and e-sports right now. In this research, Dota 2 drafting phase is studied by using stochastic processes and probabilistic models. This phase is modeled as a POMDP (Partially Observable Markov Decision Process) for the first time. Afterwards, the model is solved using data mining methods and machine learning algorithms. A large database of the game's match data is created, and six different machine learning algorithms are used to predict the match outcome based on the drafting stage. The prediction accuracy and power of these algorithms is compared, and finally, a program is developed to help players make better decisions in game, so they can increase their chances of winning
Keywords:
Stochastic Process ; Partially Observable Markov Decision Process ; Machine Learning ; Data Mining ; Artificial Neural Network ; Electronic Sports

No TOC