Εφαρμογές ταξινόμησης σε τραπεζικά δεδομένα
Master Thesis
Date
2011-05-30View/
Subject
Ταξινόμηση -- Μαθηματικά υποδείγματα ; Μεταβλητές (Μαθηματικά) ; Καταναλωτική πίστη ; Τράπεζες και τραπεζικές εργασίες -- Επεξεργασία δεδομένων ; Δάνεια ; Regression analysisAbstract
The recent Economic crisis highlighted once again the significant role of Risk monitoring of loan portfolio for Banking institutions. A primary tool for this surely is a model able to identify future potentially non-payers that could become defaulted customers. The goal for this dissertation is to describe how to build such a model for Consumer Lending, by actually constructing one, using data of a major bank of Greece. Specifically, we constructed a statistical classification model, able to predict whether at the end of the next six months, a client that currently is at 30 days past overdue, will be over 90 days overdue, or in other words whether his loan will be classified as non-performing. To serve this purpose we had, as a first step, to reduce the dimension of independent variables space. For categorical variables we applied the Two-Step Cluster Analysis, while for continuous variables Factor Analysis was chosen. In addition, Two-Step Cluster Analysis helped us in the clients’ profiling process. The main part of the analysis utilized the Logistic Regression Method, which highlighted exposure, type of loan, client’s marital status, and status of client’s business loan (if any) as the key variables for increasing customer probability to become defaulted.
Language
GreekCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 4.0 Διεθνές
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 4.0 Διεθνές