Unaprijedna neuronska mreža za procjenu kreditnog rizika

master's thesis

• Title (english): Feedforward neural network for credit risk assesment
• Author: Iva Kresnik
• Mentor: Dubravko Majetić (mentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2021-04-01, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing; Artificial Intelligence

Abstract

Cilj ovog diplomskog rada je primjena umjetnih neuronskih mreža u određivanju kreditnog rizika. U uvodu diplomskog rada objašnjena je problematika zadane teme, dana je definicija kredita te su navedeni svi programski paketi koji su korišteni u radu. U drugom poglavlju razrađena je teorija umjetnih neuronskih mreža. Zatim slijedi analiza dobivenih podataka o klijentima i kreditima od banke, kao i priprema istih za ulaz u unaprijednu neuronsku mrežu. U programskom jeziku Python 3 definirana je optimalna unaprijedna neuronska mreža pomoću paketa Keras (TensorFlow). Parametri mreže i njena struktura detaljno su opisani u četvrtom poglavlju rada, nakon čega slijedi analiza dobivenih izlaza mreže. U zaključku rada navedene su prednosti i mane korištene unaprijedne mreže, kao i moguća poboljšanja s ciljem dobivanja preciznijih predviđanja kreditnih rizika.

Abstract (english)

The aim of this thesis is to present the work of artificial neural networks application in credit score determining. In the introduction, the problems of the given assignment are explained, the definition of loan is given, and all software packages used in the paper are listed. In the second chapter, the theory of artificial neural networks is elaborated. This is followed by an analysis of the data obtained on clients and loans from the bank, as well as preprocessing the same data so it can be later used as input for the feedforward neural network. In the Python 3 programming language, the optimal feedforward neural network is defined using the Keras package (TensorFlow). The parameters of the neural network and its structure are described in detail in the fourth chapter of the paper, followed by the analysis of the obtained network outputs. The paper concludes with the advantages and disadvantages of the feedforward network that is used, as well as possible improvements in order toto obtain more accurate credit score forecasts.

• Keywords: umjetne neuronske mreže; kreditni rizik; obrada podataka
• Keywords (english): artificial neural networks; credit score; data preprocessing
• Language: croatian
• URN:NBN: urn:nbn:hr:235:451341
• Study programme: Title: Mechanical Engineering; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Embargoed access; Embargo expiration date: 2026-03-25
• Created on: 2021-03-25 10:28:34