magistrsko delo
Lea Benedičič (Author), Ljupčo Todorovski (Mentor)

Abstract

Za spletno igralnico razvijemo priporočilni sistem, ki bo igralcu na podlagi zgodovine prej izbranih iger predlagal naslednjo. Za gradnjo priporočilnega sistema uporabimo kombinacijo nenadzorovanega in nadzorovanega učenja. Nenadzorovano učenje uporabimo za hierarhično gručenje iger, kar nam pomaga rešiti problem velikega števila razpoložljivih iger v spletni igralnici. Tako se z nadzorovanim učenjem, ki uporablja prehodne matrike markovskih verig, lahko učimo verjetnosti prehodov med gručami iger. Priporočilni sistem, zgrajen na ta način, omogoča visoko točnost napovedi gruče naslednje igre, ki jo izbere igralec. Primerjava z nekaj bolj preprostimi algoritmi napovedovanja naslednje gruče potrdi premoč uporabljene kombinacije pristopov. Treba je poudariti, da razviti sistem ne priporoča posameznih iger, temveč gruče iger. Določanje posamezne igre tako prepustimo spletni igralnici, lahko pa bi v ta namen uporabili tudi v delu predstavljeno gručenje igralcev na osnovi vzorcev njihovega igranja.

Keywords

strojno učenje;gručenje podatkov;napovedno modeliranje;priporočilni sistem;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UL FU - Faculty of Administration
Publisher: [L. Benedičič]
UDC: 004.8
COBISS: 21673475 Link will open in a new window
Views: 8682
Downloads: 132
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Development of recommendation system for Web casino with machine learning
Secondary abstract: We develop a recommendation system, for the online casino, which will suggest the next game to a player based on the games previously selected. The system is based on a combination of unsupervised and supervised learning. Unsupervised learning is used for hierarchical clustering of games which helps solve the problem of large number of available games on the online casino. In turn, we use supervised learning based on Markov chain transition matrices to infer the probabilities of transition between game clusters. We empirically show that the recommendation system allows for accurate prediction of the cluster of the next game for an observed player. A comparison with simple, baseline algorithms confirms the superiority of the proposed approach. It should be emphasized that the developed system does not recommend individual games but a cluster of games. The selection of a particular game from the cluster might be left to the online casino software. Alternatively, the selection can be based on the clustering of players base on their gaming patterns.
Secondary keywords: machine learning;clustering;predictive modeling;recommendation system;
Type (COBISS): Master's thesis/paper
Study programme: 0
Embargo end date (OpenAIRE): 2021-05-15
Thesis comment: Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Finančna matematika - 2. stopnja
Pages: XI, 50 str.
ID: 11718188