Institutional Repository of Vilnius University
Būsto paskolų išankstinių grąžinimų prognozavimas naudojant mašininį mokymą /
Abstract
dc:descriptionThis work presents the application of machine learning models for the prediction of mortgage prepayment. Dataset made up of monthly observations covering the years 2005 to 2012 for anonymized Norwegian mortgage loans without contractual amortizations schemes was explored. Given the limited number of available independent variables, feature engineering was performed during which macroeconomic factors were added to the feature set. Random forest and long short-term memory artificial neural network regression models were built to predict the proportion of the outstanding notional prepaid on the individual loan level. The aforementioned predictions were aggregated by taking the average of predicted values weighted by then outstanding notional in order to obtain prepayment rate prediction on the portfolio level. Experimental evaluation of the models showed significantly improved performance of the random forest model compared to the baseline model for the time span of the training set, however a significant decline in the model’s performance was found on the forecast set. The previously mentioned issue was avoided for the long short-term memory artificial neural network model which was found to generally have the highest performance.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Martinkus, Dovydas,
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:ELABAETD192845196&prefLang=en_US