Metoda podpornih vektorjev v detekciji goljufij

magistrsko delo

Manja Krajnčič (Author), Drago Bokal (Mentor), Anja Žnidaršič (Co-mentor)

Abstract

Magistrsko delo obravnava problem odkrivanja goljufij za izbrani scenarij. Scenarij nam predstavlja eno obliko goljufanja, ki jo želimo razkriti z uporabo ustrezne metode. Kljub temu da je za odkrivanje goljufij razvitih veliko metod, pa vse niso ustrezne. Metode, ki se v prvi vrsti delijo na nadzorovane in nenadzorovane, ne odkrijejo vseh vrst goljufij, zato je zelo pomembno, da ustvarimo več scenarijev in prilagodimo metode glede na naš nabor podatkov, s tem pa pokrijemo večjo množico možnih goljufov. Za scenarij si izberemo goljufanje gostincev, nad katerim razvijemo novo metodo za odkrivanje transakcijskih goljufij. Rezultate primerjamo tudi z rezultati, ki jih nad isto množico podatkov dobimo pri uporabi metode podpornih vektorjev enega razreda. Glavni rezultat magistrskega dela predstavlja kombinacijo uporabe dveh metod za rangiranje gostincev od najbolj do najmanj sumljivih.

Keywords

magistrska dela;odkrivanje goljufij;subvencionirana študentska prehrana;metoda FSRO;metoda podpornih vektorjev enega razreda;

Data

Language:	Slovenian
Year of publishing:	2019
Typology:	2.09 - Master's Thesis
Organization:	UM FNM - Faculty of Natural Sciences and Mathematics
Publisher:	[M. Krajnčič]
UDC:	519.87(043.2)
COBISS:	24717832
Views:	862
Downloads:	67
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Support vector machine and fraud detection
Secondary abstract:	The master thesis deals with fraud detection problem for a specific scenario. The scenario represents one type of fraud that we want to detect with a proper method. There are a lot of different methods for fraud detection, but not all of them are appropriate. Methods that are classified as supervised and unsupervised, do not detect all kinds of fraud, so it is very important to create multiple scenarios in our data set to cover diverse possibilities for fraud. For our scenario, we assume that only provider can commit fraud. We then develop a new method for a fraud detection in telecommunications and compare the results with results from method one-class support vector machine. The main result of master thesis represents a combination of using these two methods for ranking providers from most to least suspicious.
Secondary keywords:	master theses;fraud detection;subsidized student meals;FSRO method;one-class support vector machine;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za naravoslovje in matematiko, Oddelek za matematiko in računalništvo
Pages:	VIII, 69 f.
ID:	11156749