Avtomatsko prepoznavanje vedenjskih vzorcev

delo diplomskega seminarja

Martin Molan (Author), Andrej Bauer (Mentor), Martin Žnidaršič (Co-mentor)

Abstract

Naloga obravnava izgradnjo napovednih modelov na osnovi realnih podatkov. Cilj opisanih postopkov je modeliranje in napovedovanje obnašanja igralcev v igralniški industriji. Osnova za izdelavo napovednega modela je priprava podatkov za algoritme strojnega učenja. Uspešnost obdelave realnih podatkov z nečistočami in napakami pomembno vpliva na možnost izgradnje smiselnega napovednega modela. Prvi cilj modeliranja vedenja je izgradnja modela, ki napoveduje, ali bo igralec naredil drugi depozit. Natančnost razvitega modela zadostuje potrebam domene in je primerna za implementacijo v praksi. Drugi cilj modeliranja je izgradnja širšega napovednega modela, ki opisuje razvoj posameznikovih igralnih navad. Predstavljena je smiselnost drugega pristopa in njegova skladnost z zahtevami domene. Preizkus napovedne moči širšega napovednega modela presega okvire te naloge saj je dolgotrajna in zahteva veliko specifičnega domenskega znanja.

Keywords

matematika;strojno učenje;nadzorovano učenje;nenadzorovano učenje;napovedno modeliranje;modeliranje vedenja;igralniška industrija;umetna inteligenca;odločitveno drevo;naključni gozd;segmentacija;

Data

Language:	Slovenian
Year of publishing:	2018
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FMF - Faculty of Mathematics and Physics
Publisher:	[M. Molan]
UDC:	004.8
COBISS:	18435161
Views:	873
Downloads:	241
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Automatic recognition of behavioral patterns
Secondary abstract:	This thesis explores development of predictive models on real life datasets. The goal of approaches, described in this thesis, is modeling and prediction of players' behavior in casino industry. The basis for creation of predictive model is preparation of real life datasets for machine learning algorithms. Sensible curation of real life datasets that include missing values, inaccuracies and other noise determines the possibility for development of accurate predictive models. First goal of predictive behavioral modeling is creation of automated prediction model that predicts if the player will make a second deposit. Accuracy of developed model is sufficient for implementation in real life casino operation. Second goal is to develop broader predictive model that describes and predicts development of player’s behavior. Sensibility of proposed approach and its compliance with domain demands is presented. Real predictive strength of proposed model is out of scope of this work as it requires a lot of additional domain knowledge.
Secondary keywords:	mathematics;machine learning;supervised learning;unsupervised learning;predictive modeling;behavioral modeling;casino industry;artificial intelligence;decision tree;random forest;clustering;
Type (COBISS):	Final seminar paper
Study programme:	0
Thesis comment:	Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Matematika - 1. stopnja
Pages:	30 str.
ID:	10959813