Title: Forecasting online job vacancy attractiveness Authors:  Miroslav Stefanik - Institute of Economic Research, Slovak Academy of Sciences (Slovakia)
Stefan Lyocsa - Slovak Academy of Sciences (Slovakia)
Zuzana Kostalova - Slovak Academy of Sciences (Slovakia) [presenting]
Abstract: The purpose is to explore whether predictions of online job vacancies (OJVs) attractiveness by job seekers, measured by i) number of job ad views, ii) response, and iii) conversion rate (the ratio of the two), could be improved. Apart from standard machine learning models, network-based feature extraction methods are used. Forecasting models utilize above 175 explanatory variables related to job characteristics, prerequisites and benefits, including simple textual features and even calendar effects. The approach could suggest what kind of data leads to the highest marginal contribution in forecasting OJV attractiveness. The findings could help employers better target prospective applicants and could be implemented in the search interface by the job portals to improve job matching, i.e., lead to improved recommender systems.