Određivanje kreditne sposobnosti klijenata korištenjem XGBoost algoritma

master's thesis

• Title (english): Determining credit default risk using the XGBoost algorithm
• Author: Josipa Škugor
• Mentor: Mate Puljiz (mentor)
• Committee member: Mate Puljiz (predsjednik povjerenstva)
• Committee member: Tomislav Burić (član povjerenstva)
• Committee member: Domagoj Vlah (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-18, Croatia

Abstract

U ovom radu istražena je primjena XGBoost algoritma za klasifikaciju klijenata Home Credit tvrtke. Cilj istraživanja je bio razviti model koji klasificira klijente na temelju povijesnih podataka i ponašanja. Algoritam koristi podatke prikupljene iz bankarskih sustava i preuzete s Kaggle stranice. Podatci su analizirani i obrađeni kako bi se prilagodili za rad s modelom. Sve analize, obrada podataka, te razvoj i implementacija modela provedeni su korištenjem Python programskog jezika. XGBoost se pokazao iznimno dobrim za rješavanje ovog problema te je treniranje samog modela, iako na velikom skupu podataka, provedeno brzo i efikasno zbog mogućnosti korištenja svih jezgri procesora. Rezultati su pokazali da bi korištenje XGBoost algoritma prilikom klasifikacije klijenata banke značajno unaprijedilo proces i olakšalo posao, a osobito u kombinaciji s još nekim dodatnim mehanizmima.

Abstract (english)

This study explores the application of the XGBoost algorithm for classifying clients of Home Credit organization. The goal of the research was to develop a model that classifies clients based on historical data and behavior. The algorithm utilizes data collected from banking systems and retrieved from Kaggle. The data were analyzed and processed to be suitably adapted for the model. All analyses, data processing, and the development and implementation of the model were carried out using the Python programming language. XGBoost proved to be exceptionally effective in addressing this problem, with the model training -- despite working with a large dataset -- being performed quickly and efficiently due to its capability to utilize all processor cores. The results indicated that using the XGBoost algorithm for client classification would significantly enhance the process and facilitate the task, especially when combined with additional mechanisms.

• Keywords: XGBoost; Home Credit; klasifikacija; Kaggle; Python
• Keywords (english): XGBoost; Home Credit; classification; Kaggle; Python
• Language: croatian
• URN:NBN: urn:nbn:hr:168:836716
• Study programme: Title: Information and Communication Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije (magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 22:17:00