Predicting Vocational Personality Type from Socio-demographic Features Using Machine Learning Methods

Abstract

This study aimed to apply supervised machine learning techniques to one domain of psychological research: vocational interests. Socio-demographic factors can be considered strong predictors of vocational interests, which might have far-reaching practical implications for professional counselling and social network analysis. The dataset used in this study is a collection of answers to the RIASEC (Holland Codes) psychological test. Different Machine Learning architectures were used to predict RIASEC scales using socio-demographic features. The problem was treated as a multioutput regression task, multiclass and multilabel classification. The following models were used: independent regression, regression chains, three-letter code classification, inferring label relations. Models comparison showed that the models that exploit intercorrelations between RIASEC scales yielded the best results.