| Title | Multinomna logistička regresija u kreditnom skoringu |
| Author | Zana Andabaka |
| Mentor(s) | Nataša Šarlija (thesis advisor)
Mirta Benšić (thesis advisor)
|
| Abstract | Jedna od metoda kvantitativne kreditne analize je logistička regresija kojom se klijent svrštava u
jednu od dvije kategorije, u dobre ili loše. Podaci o srednjim klijentima, iako postoje u bazi klijenata
svake banke, uglavnom su isključeni u procesu kreditnog skoringa. Mnoge studije do sada
nisu pokazale značajnost upotrebe srednjih klijenata pri kreiranju kredit skoring modela, uglavnom
zbog toga sto je slabija granica među definicijama dobrih, srednjih i loših, te su konačne procjene
manje točne od procjena koje bi dao logistički model u kojem su srednji izostavljeni.
U ovom radu ispitat ćemo značajnost upotrebe srednjih na bazi klijenata jedne banke, na način
da ćemo kreirati multinomni logistički model u kojem ovisna varijabla ima tri kategorije, dobre,
srednje i loše, te binomni model samo s dobrim i lošim klijentima. Te ćemo usporediti rezultate. |
| Keywords | generalized linear model multinomial distribution ordinal covariate maximum likelihood estimation credit scoring cut points |
| Granter | Josip Juraj Strossmayer University of Osijek
Department of Mathematics |
| Lower level organizational units | Chair of Pure Mathematics
Probability and Mathematical Statistics Research Group |
| Place | Osijek |
| State | Croatia |
| Scientific field, discipline, subdiscipline | NATURAL SCIENCES
Mathematics
Financial and Business Mathematics
|
| Study programme type | university |
| Study level | graduate |
| Study programme | Mathematics; specializations in: Financial and Statistical Mathematics, Mathematics and Computer Science, Industrial and Applied Mathematics |
| Study specialization | Financial and Statistical Mathematics |
| Academic title abbreviation | mag.math. |
| Genre | master's thesis |
| Language | Croatian |
| Defense date | 2017-05 |
| Parallel abstract (English) | One of quantitative credit analysis methods is logistic regression where the client is classified into
the one of two categories, good or bad.
Although there are data of poor clients in the database of any bank. They are generally excluded
in the process of credit scoring. Many studies so far haven’t shown the importance of using poor
clients in credit scoring model, mainly because the lower boundary between definitions of good,
poor and bad, so the final estimates are less accurate than estimates obtained by logistic model in
which poor are omitted.
In this article we will examine the significance of use poor clients of a bank, in a way that will
create a multinomial logistic model in which the dependent variable has three categories, good,
poor and bad, and the binomial model only with good and bad customers. Than the results will be
compared. |
| Parallel keywords (Croatian) | generalizirani linearni model multinomna distribucija ordinalna varijabla metoda maksimalne vjerodostojnosti kreditni skoring pragovi |
| Resource type | text |
| Access condition | Open access |
| Terms of use |  |
| URN:NBN | https://urn.nsk.hr/urn:nbn:hr:126:427213 |
