Institutional Repository of Vilnius University
Klasifikavimo algoritmų tyrimas ir optimizavimas vertinant kredito riziką /
Abstract
dc:descriptionClassification of the clients based on their credit risk is one of the most common problems financial institutions have to deal with. Usually, machine learning based models are applied for this task to be able to accurately classify clients into different credit risk categories. In this work, we review logistic regression, decision trees and random forest classifiers as well as their modelling appliances - data preprocessing, model evaluation and optimization; we present the results in numerical and graphical metrics. We also compare and evaluate different data balancing approaches, perform the grid-search for hyperparameters to find the most optimum configuration for maximizing AUC value. As in reality most of the data possesses some sort of noise in it affecting the classifiers in a negative way, we evaluate the proposed models’ robustness against the generated noise evaluating their ELA and AUC values with different noise levels. We propose a novel metric for evaluating noise impact on AUC value - ELAUC (equalized loss of AUC).
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Juodzevičiūtė, Giedrė,
- Contributors dc:contributor
-
- Manstavičius, Martynas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD81807544&prefLang=en_US