Institutional Repository of Vilnius University
Samdomojo darbo pajamų praleistų reikšmių užpildymas taikant daugiareikšmį įrašymą ir mašininio mokymosi metodus /
Abstract
dc:descriptionMissing values are a common problem when analysing data collected in surveys. For a variety of reasons, respondents may not answer certain questions or may provide inaccurate information, especially when the questions relate to sensitive topics such as income. The study aims to replicate a real-life situation when building a regression model to predict the income of paid employment using data collected through a survey. Missing values are common in this type of data, but both model training and forecasting require fully completed observations. Recently, machine learning methods have been increasingly used to fill in the missing values, but most studies and methods are limited to one type of variable, usually quantitative. In this study, missing value imputation is done by KNNImpute, missForest, and multiple imputation using predictive mean matching, random sampling from observed values, random forests, classification, and regression trees (CART). To predict income of paid employment, four machine learning models were employed: gradient boosting, extreme gradient boosting (XGBoost), random forests, and support vector regression (SVR). The results showed that missForest and multiple imputation with CART yield the most accurate imputations. Among the predictive models, gradient boosting delivered the best results when used with data imputed via the missForest method.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Puslytė, Guoda,
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:ELABAETD239029059&prefLang=en_US