Modeli za predviđanje ponašanja korisnika digitalne igre i odluka o plaćanju dodatnog sadržaja temeljeni na strojnom učenju

master's thesis

• Title (english): Machine Learning Based Models for Prediction of Player Behavior and Purchase Decisions in Digital Games
• Author: Filip Jurčić
• Mentor: Lea Skorin-Kapov (mentor)
• Committee member: Lea Skorin-Kapov (predsjednik povjerenstva)
• Committee member: Ognjen Dobrijević (član povjerenstva)
• Committee member: Mirko Sužnjević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-07-06, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Glavni cilj ovog rada je bio generiranje modela strojnog učenja za predviđanje odluke o plaćanju dodatnog sadražaja temeljenu na njihovom ponašanju u besplatnoj interaktivnoj storytelling igri koja je dostupna na tržištu. Objašnjena je motivacija za ovakvu vrstu istraživanja uz primjere sličnih istraživanja s rezultatima i zaključcima. Prikazan je problem pohrane i obrade ogromne količine podataka i odabrane su tehnologije i alati za rješavanje tog problema u sklopu ovog rada. Podaci o korisnicima koji su obavaili barem jednu uplatu su dodatno analizirani za određivanje vremena koje je proteklo od prve instalacije igre do kupnje (apsolutno) i vremena za kojeg je korisnik zapravo igrao igru (relativno) s grafovima za analizu distribucije korisnika. Ključni koncepti i ideja koja stoji iza machine learninga su objašnjeni s podjelom u kategorije (s nadzorom, bez nadzora) s kratkim objašnjenjem najpoznatijih algoritama za svaku kategoriju. Dan je uvod u korištenje WEKA alata, koji se koristio za klasifikaciju korsinika uz pomoć algoritama strojnog učenja. Prikazana je metoda odabira podskupa korisnika kao ulaz u algoritnme strojnog učenja u svrhu klasifikacije sa svim značajkama koje su izračunate za svakog korisnika u podskupu. Na kraju su ukratko prikazani korišteni algoritmi s detaljnim prikazom rezultata za svaki od njih.

Abstract (english)

The main goal of this thesis was the generation of machine learning models for prediction of players purchase decisions based on their behavior in a free-to-play interactive storytelling game available on the market. The motivation behind this kind of research is explained and examples of similar researches with their results and conclusions are being shown. The problem of storing and parsing huge amounts of data is presented with the technologies and tools used to solve that problem in scope of this thesis. The data for the users which have completed at least one payment has been further analyzed to determine the times of purchase in regard to the first installation of the game (absolute) and time actually spent playing (relative) with graphs to analyze the distribution of the users. The idea behind and key concepts of machine learning have been explained with a breakdown in categories (supervised, unsupervised) with brief explanation of the most popular algorithms for each of the categories. An introduction is given to the WEKA tool, which was used in classification of the users with the machine learning algorithms. The method for selecting the user subsets for the input in the machine learning algorithms have been explained, with all the features calculated for each user in the generated subset. Finally, the algorithms which have been used have been briefly introduced with detailed breakdown of the results for each of them.

• Keywords: digitalne igre; predviđanje ponašanja igrača; veliki skupovi podataka; strojno učenje
• Keywords (english): digital games; player decision prediction; big data; machine learning
• Language: english
• URN:NBN: urn:nbn:hr:168:978796
• 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: 2019-03-13 21:18:04