Institutional Repository of Vilnius University
Darbo skelbime nepateikto atlyginimo prognozavimas naudojant tekstinę skelbimo informaciją /
Abstract
dc:descriptionThis study addresses the significant issue of undisclosed salaries in job advertisements, particularly prevalent in foreign countries where salary disclosure in job listings is not mandatory by law. The primary objective was to create a data-driven classification method for predicting salaries within the IT sector of United States, thereby contributing to a more effective job search process. The proposed model also aims to help with the pay inequality, including gender-based discrepancies, by providing upfront salary expectations. Using a comprehensive dataset sourced from various US job platforms, the research explored multiple classification models — Decision Tree, Random Forest, K-Nearest Neighbors, and Gradient Boosting. To enhance model accuracy, hyperparameter optimization and different data processing techniques were tested. The study also tested hypotheses regarding the impact of such factors as company size and employee benefits on salary levels. The results demonstrated that the Gradient Boosting model outperformed others by achieving an accuracy of 81.1%, proving especially effective in predicting both the lowest and highest salary brackets. This model's integration into job search platforms could potentially improve the overall job matching process and mitigate unfair salary discrepancies. The findings also suggested that larger companies tend to offer higher salaries and that benefits like remote work and health insurance might be compensatory for lower wages. The research concluded that undisclosed salaries in job listings could unnecessarily prolong the recruitment process for both employers and employees. By applying machine learning and natural language processing techniques, the study effectively addresses this challenge within the IT job market in the US, setting a precedent for future explorations into more advanced artificial intelligence methodologies to refine the accuracy and reliability of salary predictions further.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Viaznikova, Darija,
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:ELABAETD210643879&prefLang=en_US