Detekcija žanra igre temeljena na strojnom učenju

master's thesis

• Title (english): Game genre detection based on machine learning
• Author: Tina Mustapić
• Mentor: Mirko Sužnjević (mentor)
• Committee member: Mirko Sužnjević (predsjednik povjerenstva)
• Committee member: Maja Matijašević (član povjerenstva)
• Committee member: Igor Sunday Pandžić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Cloud gaming je usluga igranja videoigri na oblaku koja radi na sličan način kao
i usluge videa na zahtjev. Videoigre se pohranjuju i izvode na daljinu na hardveru
poslužitelja, te se prenose kao video na ured̄aj igrača putem klijentskog softvera. Me-
d̄utim, cloud gaming zahtijeva pouzdanu, brzu vezu s internetom. Loša kvaliteta, kao
i zastajkivanje te veliko kašnjenje videa komplikacije su koje mogu iznimno smanjiti
iskustvenu kvalitetu igrača. Jedan od načina poboljšavanja iskustvene kvalitete bi bilo
prilagod̄avanje mrežnih postavki s obzirom na žanr videoigre. Ovaj rad opisuje tre-
niranje i validiranje 14 različitih klasifikacijskih modela na podatkovnom skupu koji
sadržava 20 videoigri podijeljenih na 5 žanrova. Model strojnog učenja koji se pokazao
kao najbolji detektor žanra videoigri, s obzirom na skupljene podatke, je Extra Trees
klasifikator, a najbolje klasificiran žanr je strategija. Najbolji rezultati su trenirani nad
cijelim skupom podataka, ali približno jednake rezultate možemo dobiti trenirajući
samo nad streaming komponentom koju je računalno jeftinije dobiti od cijelog skupa
podataka.

Abstract (english)

Cloud gaming is a cloud video game service that works in a similar way to video-
on-demand services. Video games are stored and played remotely on the server har-
dware, and are transmitted as video to the player’s device via client software. However,
cloud gaming requires a reliable, fast connection to the Internet. Poor quality, as well
as stalling and long video delays are complications that can greatly reduce the quality
of experience of a player. One way to improve the experiential quality would be to
adjust the network settings according to the video game genre. This paper describes
the training and validation of 14 different classification models on a data set containing
20 video games divided into 5 genres. The machine learning model that has proven
to be the best detector of the video game genre, given the data collected, is the Extra
Trees classifier, and the best classified genre is strategy. The best results are trained
over the whole data set, but approximately the same results can be obtained by training
only over the streaming component which is computationally cheaper to obtain than
the whole data set.

• Keywords: Strojno učenje; klasifikacija; regresija; žarn igre; video metrike; mrežne metrike; igre na računalnom oblaku
• Keywords (english): Machine learning; classification; regression; game type; video metrics; network traffic metrics; cloud gaming
• Language: croatian
• URN:NBN: urn:nbn:hr:168:187127
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 15:42:06